Search

For the efficient inference of Large Language Models (LLMs), the effective compression of key-value (KV) cache is essential. Three main types of KV cache compression techniques, namely sparsity, channel compression, and quantization, have been identified. This study presents SVDq, a Singular Value Decomposition (SVD) - based mixed precision quantization method for K cache. Initially, K cache is transformed into latent channels using SVD basis representations. Since the values in latent channels decay rapidly and become negligible after only a few latent channels, our method then incorporates importance-aware quantization and compression for latent channels. This enables the effective allocation of higher precision to more significant channels. Theoretically, we prove that SVDq results in quantization errors (x0.1 or even lower) that are much lower than those of per-channel key quantization in the original space. Our findings based on RULER and LongBench benchmarks demonstrate that SVDq can achieve an equivalent key cache precision as low as 1.25-bit. When combined with key sparsity, it can reach a key compression ratio of up to 410x for attention computation, all while maintaining comparable model performance. Notably, our method is nearly lossless for LongBench datasets. This indicates that SVDq enables high-precision low-bit quantization, providing a more efficient solution for KV cache compression in LLMs.

Audio-visual representation learning is crucial for advancing multimodal speech processing tasks, such as lipreading and audio-visual speech recognition. Recently, speech foundation models (SFMs) have shown remarkable generalization capabilities across various speech-related tasks. Building on this progress, we propose an audio-visual representation learning model that leverages cross-modal knowledge distillation from SFMs. In our method, SFMs serve as teachers, from which multi-layer hidden representations are extracted using clean audio inputs. We also introduce a multi-teacher ensemble method to distill the student, which receives audio-visual data as inputs. A novel representational knowledge distillation loss is employed to train the student during pretraining, which is also applied during finetuning to further enhance the performance on downstream tasks. Our experiments utilized both a self-supervised SFM, WavLM, and a supervised SFM, iFLYTEK-speech. The results demonstrated that our proposed method achieved superior or at least comparable performance to previous state-of-the-art baselines across automatic speech recognition, visual speech recognition, and audio-visual speech recognition tasks. Additionally, comprehensive ablation studies and the visualization of learned representations were conducted to evaluate the effectiveness of our proposed method., Comment: accepted to Pattern Recognition

The rapidly expanding Internet of Things (IoT) landscape is shifting toward cloudless architectures, removing reliance on centralized cloud services but exposing devices directly to the internet and increasing their vulnerability to cyberattacks. Our research revealed an unexpected pattern of substantial Tor network traffic targeting cloudless IoT devices. suggesting that attackers are using Tor to anonymously exploit undisclosed vulnerabilities (possibly obtained from underground markets). To delve deeper into this phenomenon, we developed TORCHLIGHT, a tool designed to detect both known and unknown threats targeting cloudless IoT devices by analyzing Tor traffic. TORCHLIGHT filters traffic via specific IP patterns, strategically deploys virtual private server (VPS) nodes for cost-effective detection, and uses a chain-of-thought (CoT) process with large language models (LLMs) for accurate threat identification. Our results are significant: for the first time, we have demonstrated that attackers are indeed using Tor to conceal their identities while targeting cloudless IoT devices. Over a period of 12 months, TORCHLIGHT analyzed 26 TB of traffic, revealing 45 vulnerabilities, including 29 zero-day exploits with 25 CVE-IDs assigned (5 CRITICAL, 3 HIGH, 16 MEDIUM, and 1 LOW) and an estimated value of approximately $312,000. These vulnerabilities affect around 12.71 million devices across 148 countries, exposing them to severe risks such as information disclosure, authentication bypass, and arbitrary command execution. The findings have attracted significant attention, sparking widespread discussion in cybersecurity circles, reaching the top 25 on Hacker News, and generating over 190,000 views., Comment: 27 pages, 14 figure, 9 tables

While Retrieval-Augmented Generation (RAG) has exhibited promise in utilizing external knowledge, its generation process heavily depends on the quality and accuracy of the retrieved context. Large language models (LLMs) struggle to evaluate the correctness of non-parametric knowledge retrieved externally when it differs from internal memorization, leading to knowledge conflicts during response generation. To this end, we introduce the Retrieval Preference Optimization (RPO), a lightweight and effective alignment method to adaptively leverage multi-source knowledge based on retrieval relevance. An implicit representation of retrieval relevance is derived and incorporated into the reward model to integrate retrieval evaluation and response generation into a single model, solving the problem that previous methods necessitate the additional procedure to assess the retrieval quality. Notably, RPO is the only RAG-dedicated alignment approach that quantifies the awareness of retrieval relevance in training, overcoming mathematical obstacles. Experiments on four datasets demonstrate that RPO outperforms RAG by 4-10% in accuracy without any extra component, exhibiting its robust generalization.

Recent advancements in personalized speech generation have brought synthetic speech increasingly close to the realism of target speakers' recordings, yet multimodal speaker generation remains on the rise. This paper introduces UniSpeaker, a unified approach for multimodality-driven speaker generation. Specifically, we propose a unified voice aggregator based on KV-Former, applying soft contrastive loss to map diverse voice description modalities into a shared voice space, ensuring that the generated voice aligns more closely with the input descriptions. To evaluate multimodality-driven voice control, we build the first multimodality-based voice control (MVC) benchmark, focusing on voice suitability, voice diversity, and speech quality. UniSpeaker is evaluated across five tasks using the MVC benchmark, and the experimental results demonstrate that UniSpeaker outperforms previous modality-specific models. Speech samples are available at \url{https://UniSpeaker.github.io}.

With the growing significance of network security, the classification of encrypted traffic has emerged as an urgent challenge. Traditional byte-based traffic analysis methods are constrained by the rigid granularity of information and fail to fully exploit the diverse correlations between bytes. To address these limitations, this paper introduces MH-Net, a novel approach for classifying network traffic that leverages multi-view heterogeneous traffic graphs to model the intricate relationships between traffic bytes. The essence of MH-Net lies in aggregating varying numbers of traffic bits into multiple types of traffic units, thereby constructing multi-view traffic graphs with diverse information granularities. By accounting for different types of byte correlations, such as header-payload relationships, MH-Net further endows the traffic graph with heterogeneity, significantly enhancing model performance. Notably, we employ contrastive learning in a multi-task manner to strengthen the robustness of the learned traffic unit representations. Experiments conducted on the ISCX and CIC-IoT datasets for both the packet-level and flow-level traffic classification tasks demonstrate that MH-Net achieves the best overall performance compared to dozens of SOTA methods., Comment: Accepted by AAAI 2025. The code is available at https://github.com/ViktorAxelsen/MH-Net. arXiv admin note: text overlap with arXiv:2402.07501

This paper proposes an Incremental Disentanglement-based Environment-Aware zero-shot text-to-speech (TTS) method, dubbed IDEA-TTS, that can synthesize speech for unseen speakers while preserving the acoustic characteristics of a given environment reference speech. IDEA-TTS adopts VITS as the TTS backbone. To effectively disentangle the environment, speaker, and text factors, we propose an incremental disentanglement process, where an environment estimator is designed to first decompose the environmental spectrogram into an environment mask and an enhanced spectrogram. The environment mask is then processed by an environment encoder to extract environment embeddings, while the enhanced spectrogram facilitates the subsequent disentanglement of the speaker and text factors with the condition of the speaker embeddings, which are extracted from the environmental speech using a pretrained environment-robust speaker encoder. Finally, both the speaker and environment embeddings are conditioned into the decoder for environment-aware speech generation. Experimental results demonstrate that IDEA-TTS achieves superior performance in the environment-aware TTS task, excelling in speech quality, speaker similarity, and environmental similarity. Additionally, IDEA-TTS is also capable of the acoustic environment conversion task and achieves state-of-the-art performance., Comment: Accepted to ICASSP 2025

Neural networks are commonly known to be vulnerable to adversarial attacks mounted through subtle perturbation on the input data. Recent development in voice-privacy protection has shown the positive use cases of the same technique to conceal speaker's voice attribute with additive perturbation signal generated by an adversarial network. This paper examines the reversibility property where an entity generating the adversarial perturbations is authorized to remove them and restore original speech (e.g., the speaker him/herself). A similar technique could also be used by an investigator to deanonymize a voice-protected speech to restore criminals' identities in security and forensic analysis. In this setting, the perturbation generative module is assumed to be known in the removal process. To this end, a joint training of perturbation generation and removal modules is proposed. Experimental results on the LibriSpeech dataset demonstrated that the subtle perturbations added to the original speech can be predicted from the anonymized speech while achieving the goal of privacy protection. By removing these perturbations from the anonymized sample, the original speech can be restored. Audio samples can be found in \url{https://voiceprivacy.github.io/Perturbation-Generation-Removal/}., Comment: 6 pages, 3 figures, published to IEEE SLT Workshop 2024

Compared to other clinical screening techniques, speech-and-language-based automated Alzheimer's disease (AD) detection methods are characterized by their non-invasiveness, cost-effectiveness, and convenience. Previous studies have demonstrated the efficacy of fine-tuning pre-trained language models (PLMs) for AD detection. However, the objective of this traditional fine-tuning method, which involves inputting only transcripts, is inconsistent with the masked language modeling (MLM) task used during the pre-training phase of PLMs. In this paper, we investigate prompt-based fine-tuning of PLMs, converting the classification task into a MLM task by inserting prompt templates into the transcript inputs. We also explore the impact of incorporating pause information from forced alignment into manual transcripts. Additionally, we compare the performance of various automatic speech recognition (ASR) models and select the Whisper model to generate ASR-based transcripts for comparison with manual transcripts. Furthermore, majority voting and ensemble techniques are applied across different PLMs (BERT and RoBERTa) using different random seeds. Ultimately, we obtain maximum detection accuracy of 95.8% (with mean 87.9%, std 3.3%) using manual transcripts, achieving state-of-the-art performance for AD detection using only transcripts on the ADReSS test set., Comment: Accepted by ISCSLP 2024

This paper proposes a novel neural denoising vocoder that can generate clean speech waveforms from noisy mel-spectrograms. The proposed neural denoising vocoder consists of two components, i.e., a spectrum predictor and a enhancement module. The spectrum predictor first predicts the noisy amplitude and phase spectra from the input noisy mel-spectrogram, and subsequently the enhancement module recovers the clean amplitude and phase spectrum from noisy ones. Finally, clean speech waveforms are reconstructed through inverse short-time Fourier transform (iSTFT). All operations are performed at the frame-level spectral domain, with the APNet vocoder and MP-SENet speech enhancement model used as the backbones for the two components, respectively. Experimental results demonstrate that our proposed neural denoising vocoder achieves state-of-the-art performance compared to existing neural vocoders on the VoiceBank+DEMAND dataset. Additionally, despite the lack of phase information and partial amplitude information in the input mel-spectrogram, the proposed neural denoising vocoder still achieves comparable performance with the serveral advanced speech enhancement methods., Comment: Accepted by NCMMSC2024

This paper proposes ESTVocoder, a novel excitation-spectral-transformed neural vocoder within the framework of source-filter theory. The ESTVocoder transforms the amplitude and phase spectra of the excitation into the corresponding speech amplitude and phase spectra using a neural filter whose backbone is ConvNeXt v2 blocks. Finally, the speech waveform is reconstructed through the inverse short-time Fourier transform (ISTFT). The excitation is constructed based on the F0: for voiced segments, it contains full harmonic information, while for unvoiced segments, it is represented by noise. The excitation provides the filter with prior knowledge of the amplitude and phase patterns, expecting to reduce the modeling difficulty compared to conventional neural vocoders. To ensure the fidelity of the synthesized speech, an adversarial training strategy is applied to ESTVocoder with multi-scale and multi-resolution discriminators. Analysis-synthesis and text-to-speech experiments both confirm that our proposed ESTVocoder outperforms or is comparable to other baseline neural vocoders, e.g., HiFi-GAN, SiFi-GAN, and Vocos, in terms of synthesized speech quality, with a reasonable model complexity and generation speed. Additional analysis experiments also demonstrate that the introduced excitation effectively accelerates the model's convergence process, thanks to the speech spectral prior information contained in the excitation., Comment: Accepted by NCMMSC2024

Assessing the naturalness of speech using mean opinion score (MOS) prediction models has positive implications for the automatic evaluation of speech synthesis systems. Early MOS prediction models took the raw waveform or amplitude spectrum of speech as input, whereas more advanced methods employed self-supervised-learning (SSL) based models to extract semantic representations from speech for MOS prediction. These methods utilized limited aspects of speech information for MOS prediction, resulting in restricted prediction accuracy. Therefore, in this paper, we propose SAMOS, a MOS prediction model that leverages both Semantic and Acoustic information of speech to be assessed. Specifically, the proposed SAMOS leverages a pretrained wav2vec2 to extract semantic representations and uses the feature extractor of a pretrained BiVocoder to extract acoustic features. These two types of features are then fed into the prediction network, which includes multi-task heads and an aggregation layer, to obtain the final MOS score. Experimental results demonstrate that the proposed SAMOS outperforms current state-of-the-art MOS prediction models on the BVCC dataset and performs comparable performance on the BC2019 dataset, according to the results of system-level evaluation metrics.

We participated in track 2 of the VoiceMOS Challenge 2024, which aimed to predict the mean opinion score (MOS) of singing samples. Our submission secured the first place among all participating teams, excluding the official baseline. In this paper, we further improve our submission and propose a novel Pitch-and-Spectrum-aware Singing Quality Assessment (PS-SQA) method. The PS-SQA is designed based on the self-supervised-learning (SSL) MOS predictor, incorporating singing pitch and spectral information, which are extracted using pitch histogram and non-quantized neural codec, respectively. Additionally, the PS-SQA introduces a bias correction strategy to address prediction biases caused by low-resource training samples, and employs model fusion technology to further enhance prediction accuracy. Experimental results confirm that our proposed PS-SQA significantly outperforms all competing systems across all system-level metrics, confirming its strong sing quality assessment capabilities.

In this paper, we propose MDCTCodec, an efficient lightweight end-to-end neural audio codec based on the modified discrete cosine transform (MDCT). The encoder takes the MDCT spectrum of audio as input, encoding it into a continuous latent code which is then discretized by a residual vector quantizer (RVQ). Subsequently, the decoder decodes the MDCT spectrum from the quantized latent code and reconstructs audio via inverse MDCT. During the training phase, a novel multi-resolution MDCT-based discriminator (MR-MDCTD) is adopted to discriminate the natural or decoded MDCT spectrum for adversarial training. Experimental results confirm that, in scenarios with high sampling rates and low bitrates, the MDCTCodec exhibited high decoded audio quality, improved training and generation efficiency, and compact model size compared to baseline codecs. Specifically, the MDCTCodec achieved a ViSQOL score of 4.18 at a sampling rate of 48 kHz and a bitrate of 6 kbps on the public VCTK corpus., Comment: Accepted by 2024 IEEE Spoken Language Technology Workshop (SLT2024)

This paper proposes a novel neural audio codec, named APCodec+, which is an improved version of APCodec. The APCodec+ takes the audio amplitude and phase spectra as the coding object, and employs an adversarial training strategy. Innovatively, we propose a two-stage joint-individual training paradigm for APCodec+. In the joint training stage, the encoder, quantizer, decoder and discriminator are jointly trained with complete spectral loss, quantization loss, and adversarial loss. In the individual training stage, the encoder and quantizer fix their parameters and provide high-quality training data for the decoder and discriminator. The decoder and discriminator are individually trained from scratch without the quantization loss. The purpose of introducing individual training is to reduce the learning difficulty of the decoder, thereby further improving the fidelity of the decoded audio. Experimental results confirm that our proposed APCodec+ at low bitrates achieves comparable performance with baseline codecs at higher bitrates, thanks to the proposed staged training paradigm., Comment: Accepted by ISCSLP 2025

The diffusion model, a new generative modeling paradigm, has achieved significant success in generating images, audio, video, and text. It has been adapted for sequence-to-sequence text generation (Seq2Seq) through DiffuSeq, termed S2S Diffusion. Existing S2S-Diffusion models predominantly rely on fixed or hand-crafted rules to schedule noise during the diffusion and denoising processes. However, these models are limited by non-contextualized noise, which fails to fully consider the characteristics of Seq2Seq tasks. In this paper, we propose the Meta-DiffuB framework - a novel scheduler-exploiter S2S-Diffusion paradigm designed to overcome the limitations of existing S2S-Diffusion models. We employ Meta-Exploration to train an additional scheduler model dedicated to scheduling contextualized noise for each sentence. Our exploiter model, an S2S-Diffusion model, leverages the noise scheduled by our scheduler model for updating and generation. Meta-DiffuB achieves state-of-the-art performance compared to previous S2S-Diffusion models and fine-tuned pre-trained language models (PLMs) across four Seq2Seq benchmark datasets. We further investigate and visualize the impact of Meta-DiffuB's noise scheduling on the generation of sentences with varying difficulties. Additionally, our scheduler model can function as a "plug-and-play" model to enhance DiffuSeq without the need for fine-tuning during the inference stage.

Current neural audio codecs typically use residual vector quantization (RVQ) to discretize speech signals. However, they often experience codebook collapse, which reduces the effective codebook size and leads to suboptimal performance. To address this problem, we introduce ERVQ, Enhanced Residual Vector Quantization, a novel enhancement strategy for the RVQ framework in neural audio codecs. ERVQ mitigates codebook collapse and boosts codec performance through both intra- and inter-codebook optimization. Intra-codebook optimization incorporates an online clustering strategy and a code balancing loss to ensure balanced and efficient codebook utilization. Inter-codebook optimization improves the diversity of quantized features by minimizing the similarity between successive quantizations. Our experiments show that ERVQ significantly enhances audio codec performance across different models, sampling rates, and bitrates, achieving superior quality and generalization capabilities. It also achieves 100% codebook utilization on one of the most advanced neural audio codecs. Further experiments indicate that audio codecs improved by the ERVQ strategy can improve unified speech-and-text large language models (LLMs). Specifically, there is a notable improvement in the naturalness of generated speech in downstream zero-shot text-to-speech tasks. Audio samples are available here.

Recent Retrieval Augmented Generation (RAG) aims to enhance Large Language Models (LLMs) by incorporating extensive knowledge retrieved from external sources. However, such approach encounters some challenges: Firstly, the original queries may not be suitable for precise retrieval, resulting in erroneous contextual knowledge; Secondly, the language model can easily generate inconsistent answer with external references due to their knowledge boundary limitation. To address these issues, we propose the chain-of-verification (CoV-RAG) to enhance the external retrieval correctness and internal generation consistency. Specifically, we integrate the verification module into the RAG, engaging in scoring, judgment, and rewriting. To correct external retrieval errors, CoV-RAG retrieves new knowledge using a revised query. To correct internal generation errors, we unify QA and verification tasks with a Chain-of-Thought (CoT) reasoning during training. Our comprehensive experiments across various LLMs demonstrate the effectiveness and adaptability compared with other strong baselines. Especially, our CoV-RAG can significantly surpass the state-of-the-art baselines using different LLM backbones., Comment: Accepted to EMNLP 2024 Findings. 9 pages, 4 figures, 7 tables

This paper proposes a novel Stage-wise and Prior-aware Neural Speech Phase Prediction (SP-NSPP) model, which predicts the phase spectrum from input amplitude spectrum by two-stage neural networks. In the initial prior-construction stage, we preliminarily predict a rough prior phase spectrum from the amplitude spectrum. The subsequent refinement stage transforms the amplitude spectrum into a refined high-quality phase spectrum conditioned on the prior phase. Networks in both stages use ConvNeXt v2 blocks as the backbone and adopt adversarial training by innovatively introducing a phase spectrum discriminator (PSD). To further improve the continuity of the refined phase, we also incorporate a time-frequency integrated difference (TFID) loss in the refinement stage. Experimental results confirm that, compared to neural network-based no-prior phase prediction methods, the proposed SP-NSPP achieves higher phase prediction accuracy, thanks to introducing the coarse phase priors and diverse training criteria. Compared to iterative phase estimation algorithms, our proposed SP-NSPP does not require multiple rounds of staged iterations, resulting in higher generation efficiency., Comment: Accepted by SLT2024

Voice anonymization has been developed as a technique for preserving privacy by replacing the speaker's voice in a speech signal with that of a pseudo-speaker, thereby obscuring the original voice attributes from machine recognition and human perception. In this paper, we focus on altering the voice attributes against machine recognition while retaining human perception. We referred to this as the asynchronous voice anonymization. To this end, a speech generation framework incorporating a speaker disentanglement mechanism is employed to generate the anonymized speech. The speaker attributes are altered through adversarial perturbation applied on the speaker embedding, while human perception is preserved by controlling the intensity of perturbation. Experiments conducted on the LibriSpeech dataset showed that the speaker attributes were obscured with their human perception preserved for 60.71% of the processed utterances., Comment: accpeted by Interspeech2024

We uncover an underlying bias present in the audio recordings produced from the picture description task of the Pitt corpus, the largest publicly accessible database for Alzheimer's Disease (AD) detection research. Even by solely utilizing the silent segments of these audio recordings, we achieve nearly 100% accuracy in AD detection. However, employing the same methods to other datasets and preprocessed Pitt recordings results in typical levels (approximately 80%) of AD detection accuracy. These results demonstrate a Clever Hans effect in AD detection on the Pitt corpus. Our findings emphasize the crucial importance of maintaining vigilance regarding inherent biases in datasets utilized for training deep learning models, and highlight the necessity for a better understanding of the models' performance., Comment: Accepted by Interspeech 2024

The majority of existing speech bandwidth extension (BWE) methods operate under the constraint of fixed source and target sampling rates, which limits their flexibility in practical applications. In this paper, we propose a multi-stage speech BWE model named MS-BWE, which can handle a set of source and target sampling rate pairs and achieve flexible extensions of frequency bandwidth. The proposed MS-BWE model comprises a cascade of BWE blocks, with each block featuring a dual-stream architecture to realize amplitude and phase extension, progressively painting the speech frequency bands stage by stage. The teacher-forcing strategy is employed to mitigate the discrepancy between training and inference. Experimental results demonstrate that our proposed MS-BWE is comparable to state-of-the-art speech BWE methods in speech quality. Regarding generation efficiency, the one-stage generation of MS-BWE can achieve over one thousand times real-time on GPU and about sixty times on CPU., Comment: Accepted by Interspeech 2024

This paper proposes a novel bidirectional neural vocoder, named BiVocoder, capable both of feature extraction and reverse waveform generation within the short-time Fourier transform (STFT) domain. For feature extraction, the BiVocoder takes amplitude and phase spectra derived from STFT as inputs, transforms them into long-frame-shift and low-dimensional features through convolutional neural networks. The extracted features are demonstrated suitable for direct prediction by acoustic models, supporting its application in text-to-speech (TTS) task. For waveform generation, the BiVocoder restores amplitude and phase spectra from the features by a symmetric network, followed by inverse STFT to reconstruct the speech waveform. Experimental results show that our proposed BiVocoder achieves better performance compared to some baseline vocoders, by comprehensively considering both synthesized speech quality and inference speed for both analysis-synthesis and TTS tasks., Comment: Accepted by Interspeech 2024

Model editing is an emerging field that focuses on updating the knowledge embedded within large language models (LLMs) without extensive retraining. However, current model editing methods significantly compromise the general abilities of LLMs as the number of edits increases, and this trade-off poses a substantial challenge to the continual learning of LLMs. In this paper, we first theoretically analyze that the factor affecting the general abilities in sequential model editing lies in the condition number of the edited matrix. The condition number of a matrix represents its numerical sensitivity, and therefore can be used to indicate the extent to which the original knowledge associations stored in LLMs are perturbed after editing. Subsequently, statistical findings demonstrate that the value of this factor becomes larger as the number of edits increases, thereby exacerbating the deterioration of general abilities. To this end, a framework termed Perturbation Restraint on Upper bouNd for Editing (PRUNE) is proposed, which applies the condition number restraints in sequential editing. These restraints can lower the upper bound on perturbation to edited models, thus preserving the general abilities. Systematically, we conduct experiments employing three popular editing methods on three LLMs across four representative downstream tasks. Evaluation results show that PRUNE can preserve considerable general abilities while maintaining the editing performance effectively in sequential model editing., Comment: Accepted by ICLR 2025

Despite recent advancements in speech generation with text prompt providing control over speech style, voice attributes in synthesized speech remain elusive and challenging to control. This paper introduces a novel task: voice attribute editing with text prompt, with the goal of making relative modifications to voice attributes according to the actions described in the text prompt. To solve this task, VoxEditor, an end-to-end generative model, is proposed. In VoxEditor, addressing the insufficiency of text prompt, a Residual Memory (ResMem) block is designed, that efficiently maps voice attributes and these descriptors into the shared feature space. Additionally, the ResMem block is enhanced with a voice attribute degree prediction (VADP) block to align voice attributes with corresponding descriptors, addressing the imprecision of text prompt caused by non-quantitative descriptions of voice attributes. We also establish the open-source VCTK-RVA dataset, which leads the way in manual annotations detailing voice characteristic differences among different speakers. Extensive experiments demonstrate the effectiveness and generalizability of our proposed method in terms of both objective and subjective metrics. The dataset and audio samples are available on the website.

This paper presents a novel neural speech phase prediction model which predicts wrapped phase spectra directly from amplitude spectra. The proposed model is a cascade of a residual convolutional network and a parallel estimation architecture. The parallel estimation architecture is a core module for direct wrapped phase prediction. This architecture consists of two parallel linear convolutional layers and a phase calculation formula, imitating the process of calculating the phase spectra from the real and imaginary parts of complex spectra and strictly restricting the predicted phase values to the principal value interval. To avoid the error expansion issue caused by phase wrapping, we design anti-wrapping training losses defined between the predicted wrapped phase spectra and natural ones by activating the instantaneous phase error, group delay error and instantaneous angular frequency error using an anti-wrapping function. We mathematically demonstrate that the anti-wrapping function should possess three properties, namely parity, periodicity and monotonicity. We also achieve low-latency streamable phase prediction by combining causal convolutions and knowledge distillation training strategies. For both analysis-synthesis and specific speech generation tasks, experimental results show that our proposed neural speech phase prediction model outperforms the iterative phase estimation algorithms and neural network-based phase prediction methods in terms of phase prediction precision, efficiency and robustness. Compared with HiFi-GAN-based waveform reconstruction method, our proposed model also shows outstanding efficiency advantages while ensuring the quality of synthesized speech. To the best of our knowledge, we are the first to directly predict speech phase spectra from amplitude spectra only via neural networks., Comment: Accepted by IEEE Transactions on Audio, Speech and Language Processing. arXiv admin note: substantial text overlap with arXiv:2211.15974

Audio-text retrieval (ATR), which retrieves a relevant caption given an audio clip (A2T) and vice versa (T2A), has recently attracted much research attention. Existing methods typically aggregate information from each modality into a single vector for matching, but this sacrifices local details and can hardly capture intricate relationships within and between modalities. Furthermore, current ATR datasets lack comprehensive alignment information, and simple binary contrastive learning labels overlook the measurement of fine-grained semantic differences between samples. To counter these challenges, we present a novel ATR framework that comprehensively captures the matching relationships of multimodal information from different perspectives and finer granularities. Specifically, a fine-grained alignment method is introduced, achieving a more detail-oriented matching through a multiscale process from local to global levels to capture meticulous cross-modal relationships. In addition, we pioneer the application of cross-modal similarity consistency, leveraging intra-modal similarity relationships as soft supervision to boost more intricate alignment. Extensive experiments validate the effectiveness of our approach, outperforming previous methods by significant margins of at least 3.9% (T2A) / 6.9% (A2T) R@1 on the AudioCaps dataset and 2.9% (T2A) / 5.4% (A2T) R@1 on the Clotho dataset., Comment: 5 pages, accepted to ICASSP2024

This paper introduces a novel neural audio codec targeting high waveform sampling rates and low bitrates named APCodec, which seamlessly integrates the strengths of parametric codecs and waveform codecs. The APCodec revolutionizes the process of audio encoding and decoding by concurrently handling the amplitude and phase spectra as audio parametric characteristics like parametric codecs. It is composed of an encoder and a decoder with the modified ConvNeXt v2 network as the backbone, connected by a quantizer based on the residual vector quantization (RVQ) mechanism. The encoder compresses the audio amplitude and phase spectra in parallel, amalgamating them into a continuous latent code at a reduced temporal resolution. This code is subsequently quantized by the quantizer. Ultimately, the decoder reconstructs the audio amplitude and phase spectra in parallel, and the decoded waveform is obtained by inverse short-time Fourier transform. To ensure the fidelity of decoded audio like waveform codecs, spectral-level loss, quantization loss, and generative adversarial network (GAN) based loss are collectively employed for training the APCodec. To support low-latency streamable inference, we employ feed-forward layers and causal deconvolutional layers in APCodec, incorporating a knowledge distillation training strategy to enhance the quality of decoded audio. Experimental results confirm that our proposed APCodec can encode 48 kHz audio at bitrate of just 6 kbps, with no significant degradation in the quality of the decoded audio. At the same bitrate, our proposed APCodec also demonstrates superior decoded audio quality and faster generation speed compared to well-known codecs, such as Encodec, AudioDec and DAC., Comment: Published at IEEE/ACM Transactions on Audio, Speech, and Language Processing

As network security receives widespread attention, encrypted traffic classification has become the current research focus. However, existing methods conduct traffic classification without sufficiently considering the common characteristics between data samples, leading to suboptimal performance. Moreover, they train the packet-level and flow-level classification tasks independently, which is redundant because the packet representations learned in the packet-level task can be exploited by the flow-level task. Therefore, in this paper, we propose an effective model named a Contrastive Learning Enhanced Temporal Fusion Encoder (CLE-TFE). In particular, we utilize supervised contrastive learning to enhance the packet-level and flow-level representations and perform graph data augmentation on the byte-level traffic graph so that the fine-grained semantic-invariant characteristics between bytes can be captured through contrastive learning. We also propose cross-level multi-task learning, which simultaneously accomplishes the packet-level and flow-level classification tasks in the same model with one training. Further experiments show that CLE-TFE achieves the best overall performance on the two tasks, while its computational overhead (i.e., floating point operations, FLOPs) is only about 1/14 of the pre-trained model (e.g., ET-BERT). We release the code at https://github.com/ViktorAxelsen/CLE-TFE, Comment: The code is available at https://github.com/ViktorAxelsen/CLE-TFE

Despite their exceptional capabilities, large language models (LLMs) are prone to generating unintended text due to false or outdated knowledge. Given the resource-intensive nature of retraining LLMs, there has been a notable increase in the development of knowledge editing. However, current approaches and evaluations rarely explore the perturbation of editing on neighboring knowledge. This paper studies whether updating new knowledge to LLMs perturbs the neighboring knowledge encapsulated within them. Specifically, we seek to figure out whether appending a new answer into an answer list to a factual question leads to catastrophic forgetting of original correct answers in this list, as well as unintentional inclusion of incorrect answers. A metric of additivity is introduced and a benchmark dubbed as Perturbation Evaluation of Appending Knowledge (PEAK) is constructed to evaluate the degree of perturbation to neighboring knowledge when appending new knowledge. Besides, a plug-and-play framework termed Appending via Preservation and Prevention (APP) is proposed to mitigate the neighboring perturbation by maintaining the integrity of the answer list. Experiments demonstrate the effectiveness of APP coupling with four editing methods on four LLMs. The code and data are available at https://github.com/mjy1111/PEAK., Comment: Accepted by ICML 2024

Large language models (LLMs) inevitably exhibit hallucinations since the accuracy of generated texts cannot be secured solely by the parametric knowledge they encapsulate. Although retrieval-augmented generation (RAG) is a practicable complement to LLMs, it relies heavily on the relevance of retrieved documents, raising concerns about how the model behaves if retrieval goes wrong. To this end, we propose the Corrective Retrieval Augmented Generation (CRAG) to improve the robustness of generation. Specifically, a lightweight retrieval evaluator is designed to assess the overall quality of retrieved documents for a query, returning a confidence degree based on which different knowledge retrieval actions can be triggered. Since retrieval from static and limited corpora can only return sub-optimal documents, large-scale web searches are utilized as an extension for augmenting the retrieval results. Besides, a decompose-then-recompose algorithm is designed for retrieved documents to selectively focus on key information and filter out irrelevant information in them. CRAG is plug-and-play and can be seamlessly coupled with various RAG-based approaches. Experiments on four datasets covering short- and long-form generation tasks show that CRAG can significantly improve the performance of RAG-based approaches., Comment: Update results, add more analysis, and fix typos

Speech bandwidth extension (BWE) refers to widening the frequency bandwidth range of speech signals, enhancing the speech quality towards brighter and fuller. This paper proposes a generative adversarial network (GAN) based BWE model with parallel prediction of Amplitude and Phase spectra, named AP-BWE, which achieves both high-quality and efficient wideband speech waveform generation. The proposed AP-BWE generator is entirely based on convolutional neural networks (CNNs). It features a dual-stream architecture with mutual interaction, where the amplitude stream and the phase stream communicate with each other and respectively extend the high-frequency components from the input narrowband amplitude and phase spectra. To improve the naturalness of the extended speech signals, we employ a multi-period discriminator at the waveform level and design a pair of multi-resolution amplitude and phase discriminators at the spectral level, respectively. Experimental results demonstrate that our proposed AP-BWE achieves state-of-the-art performance in terms of speech quality for BWE tasks targeting sampling rates of both 16 kHz and 48 kHz. In terms of generation efficiency, due to the all-convolutional architecture and all-frame-level operations, the proposed AP-BWE can generate 48 kHz waveform samples 292.3 times faster than real-time on a single RTX 4090 GPU and 18.1 times faster than real-time on a single CPU. Notably, to our knowledge, AP-BWE is the first to achieve the direct extension of the high-frequency phase spectrum, which is beneficial for improving the effectiveness of existing BWE methods., Comment: Accepted by IEEE/ACM Transactions on Audio, Speech, and Language Processing

Model editing is a technique that edits the large language models (LLMs) with updated knowledge to alleviate hallucinations without resource-intensive retraining. While current model editing methods can effectively modify a model's behavior within a specific area of interest, they often overlook the potential unintended side effects on the general abilities of LLMs such as reasoning, natural language inference, and question answering. In this paper, we raise concerns that model editing's improvements on factuality may come at the cost of a significant degradation of the model's general abilities. We systematically analyze the side effects by evaluating four popular editing methods on three LLMs across eight representative tasks. Our extensive empirical experiments show that it is challenging for current editing methods to simultaneously improve factuality of LLMs and maintain their general abilities. Our analysis reveals that the side effects are caused by model editing altering the original model weights excessively, leading to overfitting to the edited facts. To mitigate this, a method named RECT is proposed to regularize the edit update weights by imposing constraints on their complexity based on the RElative Change in weighT. Evaluation results show that RECT can significantly mitigate the side effects of editing while still maintaining over 94% editing performance., Comment: Accepted by EMNLP 2024

While several long-form VideoQA datasets have been introduced, the length of both videos used to curate questions and sub-clips of clues leveraged to answer those questions have not yet reached the criteria for genuine long-form video understanding. Moreover, their QAs are unduly narrow and modality-biased, lacking a wider view of understanding long-term video content with rich dynamics and complex narratives. To remedy this, we introduce MoVQA, a long-form movie question-answering dataset, and benchmark to assess the diverse cognitive capabilities of multimodal systems rely on multi-level temporal lengths, with considering both video length and clue length. Additionally, to take a step towards human-level understanding in long-form video, versatile and multimodal question-answering is designed from the moviegoer-perspective to assess the model capabilities on various perceptual and cognitive axes.Through analysis involving various baselines reveals a consistent trend: the performance of all methods significantly deteriorate with increasing video and clue length. Meanwhile, our established baseline method has shown some improvements, but there is still ample scope for enhancement on our challenging MoVQA dataset. We expect our MoVQA to provide a new perspective and encourage inspiring works on long-form video understanding research.

Speech enhancement is widely used as a front-end to improve the speech quality in many audio systems, while it is hard to extract the target speech in multi-talker conditions without prior information on the speaker identity. It was shown that the auditory attention on the target speaker can be decoded from the electroencephalogram (EEG) of the listener implicitly. In this work, we therefore propose a novel end-to-end brain-assisted speech enhancement network (BASEN), which incorporates the listeners' EEG signals and adopts a temporal convolutional network together with a convolutional multi-layer cross attention module to fuse EEG-audio features. Considering that an EEG cap with sparse channels exhibits multiple benefits and in practice many electrodes might contribute marginally, we further propose two channel selection methods, called residual Gumbel selection and convolutional regularization selection. They are dedicated to tackling training instability and duplicated channel selections, respectively. Experimental results on a public dataset show the superiority of the proposed BASEN over existing approaches. The proposed channel selection methods can significantly reduce the amount of informative EEG channels with a negligible impact on the performance., Comment: arXiv admin note: text overlap with arXiv:2305.09994

In our previous work, we proposed a neural vocoder called APNet, which directly predicts speech amplitude and phase spectra with a 5 ms frame shift in parallel from the input acoustic features, and then reconstructs the 16 kHz speech waveform using inverse short-time Fourier transform (ISTFT). APNet demonstrates the capability to generate synthesized speech of comparable quality to the HiFi-GAN vocoder but with a considerably improved inference speed. However, the performance of the APNet vocoder is constrained by the waveform sampling rate and spectral frame shift, limiting its practicality for high-quality speech synthesis. Therefore, this paper proposes an improved iteration of APNet, named APNet2. The proposed APNet2 vocoder adopts ConvNeXt v2 as the backbone network for amplitude and phase predictions, expecting to enhance the modeling capability. Additionally, we introduce a multi-resolution discriminator (MRD) into the GAN-based losses and optimize the form of certain losses. At a common configuration with a waveform sampling rate of 22.05 kHz and spectral frame shift of 256 points (i.e., approximately 11.6ms), our proposed APNet2 vocoder outperformed the original APNet and Vocos vocoders in terms of synthesized speech quality. The synthesized speech quality of APNet2 is also comparable to that of HiFi-GAN and iSTFTNet, while offering a significantly faster inference speed.

Large Language Models (LLMs) have emerged as influential instruments within the realm of natural language processing; nevertheless, their capacity to handle multi-party conversations (MPCs) -- a scenario marked by the presence of multiple interlocutors involved in intricate information exchanges -- remains uncharted. In this paper, we delve into the potential of generative LLMs such as ChatGPT and GPT-4 within the context of MPCs. An empirical analysis is conducted to assess the zero-shot learning capabilities of ChatGPT and GPT-4 by subjecting them to evaluation across three MPC datasets that encompass five representative tasks. The findings reveal that ChatGPT's performance on a number of evaluated MPC tasks leaves much to be desired, whilst GPT-4's results portend a promising future. Additionally, we endeavor to bolster performance through the incorporation of MPC structures, encompassing both speaker and addressee architecture. This study provides an exhaustive evaluation and analysis of applying generative LLMs to MPCs, casting a light upon the conception and creation of increasingly effective and robust MPC agents. Concurrently, this work underscores the challenges implicit in the utilization of LLMs for MPCs, such as deciphering graphical information flows and generating stylistically consistent responses., Comment: Accepted by Findings of EMNLP 2023

Recent studies have demonstrated that large language models (LLMs) store massive factual knowledge within their parameters. But existing LLMs are prone to hallucinate unintended text due to false or outdated knowledge. Since retraining LLMs is resource intensive, there has been a growing interest in the concept of model editing. Despite the emergence of benchmarks and approaches, these unidirectional editing and evaluation have failed to explore the reversal curse. Intuitively, if "The capital of France is" is edited to be a counterfact "London" within a model, then it should be able to naturally reason and recall the reverse fact, i.e., "London is the capital of" followed by "France" instead of "England". In this paper, we study bidirectional language model editing, aiming to provide rigorous model editing evaluation to assess if edited LLMs can recall the editing knowledge bidirectionally. A new evaluation metric of reversibility is introduced, and a benchmark dubbed as Bidirectional Assessment for Knowledge Editing (BAKE) is constructed to evaluate the reversibility of edited models in recalling knowledge in the reverse direction of editing. We surprisingly observe that while current editing methods and LLMs can effectively recall editing facts in the direction of editing, they suffer serious deficiencies when evaluated in the reverse direction. To mitigate the reversal curse, a method named Bidirectionally Inversible Relationship moDeling (BIRD) is proposed. A set of editing objectives that incorporate bidirectional relationships between subject and object into the updated model weights are designed. Experiments show that BIRD improves the performance of four representative LLMs of different sizes via question answering and judgement.

Audio-visual speech enhancement (AV-SE) aims to enhance degraded speech along with extra visual information such as lip videos, and has been shown to be more effective than audio-only speech enhancement. This paper proposes the incorporation of ultrasound tongue images to improve the performance of lip-based AV-SE systems further. To address the challenge of acquiring ultrasound tongue images during inference, we first propose to employ knowledge distillation during training to investigate the feasibility of leveraging tongue-related information without directly inputting ultrasound tongue images. Specifically, we guide an audio-lip speech enhancement student model to learn from a pre-trained audio-lip-tongue speech enhancement teacher model, thus transferring tongue-related knowledge. To better model the alignment between the lip and tongue modalities, we further propose the introduction of a lip-tongue key-value memory network into the AV-SE model. This network enables the retrieval of tongue features based on readily available lip features, thereby assisting the subsequent speech enhancement task. Experimental results demonstrate that both methods significantly improve the quality and intelligibility of the enhanced speech compared to traditional lip-based AV-SE baselines. Moreover, both proposed methods exhibit strong generalization performance on unseen speakers and in the presence of unseen noises. Furthermore, phone error rate (PER) analysis of automatic speech recognition (ASR) reveals that while all phonemes benefit from introducing ultrasound tongue images, palatal and velar consonants benefit most., Comment: Submmited to IEEE/ACM Transactions on Audio, Speech and Language Processing. arXiv admin note: text overlap with arXiv:2305.14933

This paper presents a novel task, zero-shot voice conversion based on face images (zero-shot FaceVC), which aims at converting the voice characteristics of an utterance from any source speaker to a newly coming target speaker, solely relying on a single face image of the target speaker. To address this task, we propose a face-voice memory-based zero-shot FaceVC method. This method leverages a memory-based face-voice alignment module, in which slots act as the bridge to align these two modalities, allowing for the capture of voice characteristics from face images. A mixed supervision strategy is also introduced to mitigate the long-standing issue of the inconsistency between training and inference phases for voice conversion tasks. To obtain speaker-independent content-related representations, we transfer the knowledge from a pretrained zero-shot voice conversion model to our zero-shot FaceVC model. Considering the differences between FaceVC and traditional voice conversion tasks, systematic subjective and objective metrics are designed to thoroughly evaluate the homogeneity, diversity and consistency of voice characteristics controlled by face images. Through extensive experiments, we demonstrate the superiority of our proposed method on the zero-shot FaceVC task. Samples are presented on our demo website.

Phase information has a significant impact on speech perceptual quality and intelligibility. However, existing speech enhancement methods encounter limitations in explicit phase estimation due to the non-structural nature and wrapping characteristics of the phase, leading to a bottleneck in enhanced speech quality. To overcome the above issue, in this paper, we proposed MP-SENet, a novel Speech Enhancement Network that explicitly enhances Magnitude and Phase spectra in parallel. The proposed MP-SENet comprises a Transformer-embedded encoder-decoder architecture. The encoder aims to encode the input distorted magnitude and phase spectra into time-frequency representations, which are further fed into time-frequency Transformers for alternatively capturing time and frequency dependencies. The decoder comprises a magnitude mask decoder and a phase decoder, directly enhancing magnitude and wrapped phase spectra by incorporating a magnitude masking architecture and a phase parallel estimation architecture, respectively. Multi-level loss functions explicitly defined on the magnitude spectra, wrapped phase spectra, and short-time complex spectra are adopted to jointly train the MP-SENet model. A metric discriminator is further employed to compensate for the incomplete correlation between these losses and human auditory perception. Experimental results demonstrate that our proposed MP-SENet achieves state-of-the-art performance across multiple speech enhancement tasks, including speech denoising, dereverberation, and bandwidth extension. Compared to existing phase-aware speech enhancement methods, it further mitigates the compensation effect between the magnitude and phase by explicit phase estimation, elevating the perceptual quality of enhanced speech., Comment: Submmited to IEEE Transactions on Audio, Speech and Language Processing

Speech phase prediction, which is a significant research focus in the field of signal processing, aims to recover speech phase spectra from amplitude-related features. However, existing speech phase prediction methods are constrained to recovering phase spectra with short frame shifts, which are considerably smaller than the theoretical upper bound required for exact waveform reconstruction of short-time Fourier transform (STFT). To tackle this issue, we present a novel long-frame-shift neural speech phase prediction (LFS-NSPP) method which enables precise prediction of long-frame-shift phase spectra from long-frame-shift log amplitude spectra. The proposed method consists of three stages: interpolation, prediction and decimation. The short-frame-shift log amplitude spectra are first constructed from long-frame-shift ones through frequency-by-frequency interpolation to enhance the spectral continuity, and then employed to predict short-frame-shift phase spectra using an NSPP model, thereby compensating for interpolation errors. Ultimately, the long-frame-shift phase spectra are obtained from short-frame-shift ones through frame-by-frame decimation. Experimental results show that the proposed LFS-NSPP method can yield superior quality in predicting long-frame-shift phase spectra than the original NSPP model and other signal-processing-based phase estimation algorithms., Comment: Published at IEEE Signal Processing Letters

Audio-visual speech enhancement (AV-SE) aims to enhance degraded speech along with extra visual information such as lip videos, and has been shown to be more effective than audio-only speech enhancement. This paper proposes further incorporating ultrasound tongue images to improve lip-based AV-SE systems' performance. Knowledge distillation is employed at the training stage to address the challenge of acquiring ultrasound tongue images during inference, enabling an audio-lip speech enhancement student model to learn from a pre-trained audio-lip-tongue speech enhancement teacher model. Experimental results demonstrate significant improvements in the quality and intelligibility of the speech enhanced by the proposed method compared to the traditional audio-lip speech enhancement baselines. Further analysis using phone error rates (PER) of automatic speech recognition (ASR) shows that palatal and velar consonants benefit most from the introduction of ultrasound tongue images., Comment: Published in InterSpeech 2023

This paper proposes MP-SENet, a novel Speech Enhancement Network which directly denoises Magnitude and Phase spectra in parallel. The proposed MP-SENet adopts a codec architecture in which the encoder and decoder are bridged by convolution-augmented transformers. The encoder aims to encode time-frequency representations from the input noisy magnitude and phase spectra. The decoder is composed of parallel magnitude mask decoder and phase decoder, directly recovering clean magnitude spectra and clean-wrapped phase spectra by incorporating learnable sigmoid activation and parallel phase estimation architecture, respectively. Multi-level losses defined on magnitude spectra, phase spectra, short-time complex spectra, and time-domain waveforms are used to train the MP-SENet model jointly. Experimental results show that our proposed MP-SENet achieves a PESQ of 3.50 on the public VoiceBank+DEMAND dataset and outperforms existing advanced speech enhancement methods., Comment: Accepted by Interspeech 2023