Search

ObjectiveAffective disorders are associated with atypical voice patterns; however, automated voice analyses suffer from small sample sizes and untested generalizability on external data. We investigated a generalizable approach to aid clinical evaluation of depression and remission from voice using transfer learning: we train machine learning models on easily accessible non-clinical datasets and test them on novel clinical data in a different language.MethodsA Mixture-of-Experts machine learning model was trained to infer happy/sad emotional state using three publicly available emotional speech corpora in German and US English. We examined the model’s predictive ability to classify the presence of depression on Danish speaking healthy controls (N = 42), patients with first-episode major depressive disorder (MDD) (N = 40), and the subset of the same patients who entered remission (N = 25) based on recorded clinical interviews. The model was evaluated on raw, de-noised, and speaker-diarized data.ResultsThe model showed separation between healthy controls and depressed patients at the first visit, obtaining an AUC of 0.71. Further, speech from patients in remission was indistinguishable from that of the control group. Model predictions were stable throughout the interview, suggesting that 20-30 seconds of speech might be enough to accurately screen a patient. Background noise (but not speaker diarization) heavily impacted predictions.ConclusionA generalizable speech emotion recognition model can effectively reveal changes in speaker depressive states before and after remission in patients with MDD. Data collection settings and data cleaning are crucial when considering automated voice analysis for clinical purposes.Significant outcomes- Using a speech emotion recognition model trained on other languages, we predicted the presence of MDD with an AUC of 0.71.- The speech emotion recognition model could accurately detect changes in voice after patients achieved remission from MDD.- Preprocessing steps, particularly background noise removal, greatly influenced classification performance.Limitations- No data from non-remitters, meaning that changes to voice for that group could not be assessed.- It is unclear how well the model would generalize beyond Germanic languages.Data availability statementDue to the nature of the data (autobiographical interviews in a clinical population), the recordings of the participants cannot be shared publicly. The aggregated model predictions and code used to run the analyses is available athttps://github.com/HLasse/SERDepressionDetection.

Simultaneous speech of multiple speakers is known as overlapped speech, which causes problems for speech recognition and speaker diarization systems. The present work uses previously less utilized signal phase information in the task of overlapped speech detection. In this context, Instantaneous Frequency Cosine Coefficient (IFCC) and Modified Group Delay Cepstral Coefficient (MGDCC) features are explored. IFCC captures the time-varying phase characteristics, while MGDCC represents the frequency-varying information of the phase spectrum. A Convolutional Neural Network and Long Short-Term Memory (CNN-LSTM)-based classifier is used for the classification. The present work uses synthetically generated overlapped speech from the GRID corpus. The proposed method is benchmarked against three baseline approaches that use magnitude spectrum features. It is observed that the combination of IFCC and MGDCC features with CNN-LSTM classifier provides better performance than the baselines. The combination of phase features with magnitude-based MFCC feature provides the best performance, indicating the importance of complementary information. The present study also investigates the effect of segment duration, genders, and number of simultaneous speakers on the overlapped speech detection system. Finally, the proposed method is also evaluated on real overlapped data from the AMI corpus.

Speaker recognition is a task of identifying persons from their voices. Recently, deep learning has dramatically revolutionized speaker recognition. However, there is lack of comprehensive reviews on the exciting progress. In this paper, we review several major subtasks of speaker recognition, including speaker verification, identification, diarization, and robust speaker recognition, with a focus on deep-learning-based methods. Because the major advantage of deep learning over conventional methods is its representation ability, which is able to produce highly abstract embedding features from utterances, we first pay close attention to deep-learning-based speaker feature extraction, including the inputs, network structures, temporal pooling strategies, and objective functions respectively, which are the fundamental components of many speaker recognition subtasks. Then, we make an overview of speaker diarization, with an emphasis of recent supervised, end-to-end, and online diarization. Finally, we survey robust speaker recognition from the perspectives of domain adaptation and speech enhancement, which are two major approaches of dealing with domain mismatch and noise problems. Popular and recently released corpora are listed at the end of the paper.

Humans can identify a speaker by listening to their voice, over the telephone, or on any digital devices. Acquiring this congenital human competency, authentication technologies based on voice biometrics, such as automatic speaker recognition (ASR), have been introduced. An ASR recognizes speakers by analyzing speech signals and characteristics extracted from speaker’s voices. ASR has recently become an effective research area as an essential aspect of voice biometrics. Specifically, this literature survey gives a concise introduction to ASR and provides an overview of the general architectures dealing with speaker recognition technologies, and upholds the past, present, and future research trends in this area. This paper briefly describes all the main aspects of ASR, such as speaker identification, verification, diarization etc. Further, the performance of current speaker recognition systems are investigated in this survey with the limitations and possible ways of improvement. Finally, a few unsolved challenges of speaker recognition are presented at the closure of this survey.

Speaker role recognition is an important component in multimedia analysis for applications such as speaker naming, speaker diarization and video summarization. The lack of labeled datasets for this task has constrained algorithm evaluations. In this paper, we present a new multimodal dataset for speaker role recognition in Arabic TV programs. The dataset is artificially created using data provided by the Multi-Genre Broadcast challenge dataset. We also describe our algorithm for the processing and creation of speaker segments and their corresponding transcripts from audio documents. The spoken transcript and the speaker segments are automatically annotated for their speaker role of presenter, reporter, or a guest speaker. Based on these artificial annotations, we demonstrate for the speaker role labeling the importance of taking into account multimodal information for predicting speaker role. We present a monomodal and multimodal speaker role recognition approaches on speaker segments mined from television programs, with audio and textual classification baselines over a three-way speaker role labeling of presenter, reporter and guest.

Speech technology systems such as Automatic Speech Recognition (ASR), speaker diarization, speaker recognition, and speech synthesis have advanced significantly by the emergence of deep learning techniques. However, none of these voice-enabled systems perform well in natural environmental circumstances, specifically in situations where one or more potential interfering talkers are involved. Therefore, overlapping speech detection has become an important front-end triage step for speech technology applications. This is crucial for large-scale datsets where manual labeling in not possible. A block-based CNN architecture is proposed to address modeling overlapping speech in audio streams with frames as short as 25 ms. The proposed architecture is robust to both: (i) shifts in distribution of network activations due to the change in network parameters during training, (ii) local variations from the input features caused by feature extraction, environmental noise, or room interference. We also investigate the effect of alternate input features including spectral magnitude, MFCC, MFB, and pyknogram on both computational time and classification performance. Evaluation is performed on simulated overlapping speech signals based on the GRID corpus. The experimental results highlight the capability of the proposed system in detecting overlapping speech frames with 90.5% accuracy, 93.5% precision, 92.7% recall, and 92.8% Fscore on same gender overlapped speech. For opposite gender cases, the network scores exceed 95% in all the classification metrics.

The performance of most speaker diarization systems with x-vector embeddings is both vulnerable to noisy environments and lacks domain robustness. Earlier work on speaker diarization using generative adversarial network (GAN) with an encoder network (ClusterGAN) to project input x-vectors into a latent space has shown promising performance on meeting data. In this paper, we extend the ClusterGAN network to improve diarization robustness and enable rapid generalization across various challenging domains. To this end, we fetch the pre-trained encoder from the ClusterGAN and fine-tune it by using prototypical loss (meta-ClusterGAN or MCGAN) under the meta-learning paradigm. Experiments are conducted on CALLHOME telephonic conversations, AMI meeting data, DIHARD II (dev set) which includes challenging multi-domain corpus, and two child-clinician interaction corpora (ADOS, BOSCC) related to the autism spectrum disorder domain. Extensive analyses of the experimental data are done to investigate the effectiveness of the proposed ClusterGAN and MCGAN embeddings over x-vectors. The results show that the proposed embeddings with normalized maximum eigengap spectral clustering (NME-SC) back-end consistently outperform Kaldi state-of-the-art z-vector diarization system. Finally, we employ embedding fusion with x-vectors to provide further improvement in diarization performance. We achieve a relative diarization error rate (DER) improvement of 6.67% to 53.93% on the aforementioned datasets using the proposed fused embeddings over x-vectors. Besides, the MCGAN embeddings provide better performance in the number of speakers estimation and short speech segment diarization as compared to x-vectors and ClusterGAN in telephonic data., Comment: Submitted to IEEE/ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING

The diarization of the self emerged through historically contingent social and economic relations and the practice of accounting as an integral aspect of industrialization and capitalism. From the journal to new media technologies, the recording of everyday life is about the governance and construction of the self, illuminating the ‘everyday’ as a salient dimension of sense-making in relation to the wider world. Accounting the self today has gained renewed prominence through new media technologies and social media platforms and its increasing incorporation into everyday life, inviting a public interface. The teleology of older forms of accounting the self (i.e. the journal) to new media technologies needs to be seen through a continuum in which sociocultural and economic forces enmesh with the medium of communication as a tool of self-expression to capture the self in its diurnal mode. This article utilizes Raymond Williams’ notion of cultural materialism, in linking the diary as a modern technology of self-expression with vlogs in the post-digital terrain wherein material practices shed light onto our lifeworlds construed through the resources available within material and immaterial infrastructures, offering both agency and abstraction by capital. As such, the manifest cultural form is intimately mediated by cultural technologies.

Speaker indexing or diarization is the process of automatically partitioning the conversation involving multiple speakers into homogeneous segments and grouping together all the segments that correspond to the same speaker. So far, certain works have been done under this aspect; still, the need of accurate partitioning process gets lagged under certain criteria. With this in mind, this paper aims to introduce a new speaker indexing or diarization model (Telugu language) that initially involves Mel Frequency Cepstral coefficient based feature extraction. Subsequently, a new Optimized Artificial Neural Network (ANN) is introduced for clustering process. The novelty behind the clustering process is: the training of ANN takes place through optimization logic that updates the weight of ANN by a hybrid concept of Artificial Bee Colony (ABC) and Lion Algorithm (LA). Thereby, the proposed model is named as ANN-ABC-LA model. Finally, the performance of the proposed ANN-ABC-LA model is compared over the state-of-the-art models with respect to different performance measures.

Given the limited accessibility to Speech and Language Pathologists (SLPs) children in need often have, pediatric Computer-Aided Speech Therapy (CAST) tools can play an important role in the early diagnosis and treatment of speech disorders. However, various challenges impede the implementation of accurate automated analysis of speech disorders in children. In this article, we first discuss three key challenges in processing child disordered speech: 1) the unreliability of low-level annotation and scarcity of speech corpora, 2) speaker diarization of therapy sessions and 3) inaccurate children's acoustic models. We next explore opportunities to overcome some of these challenges. First, we investigate the effectiveness of high-level paralinguistic features in disordered speech detection to reduce the dependency on annotated data. A binary classifier trained using paralinguistic features extracted from both typically developing children and those suffering from Speech Sound Disorders (SSD) achieved 87% subject-level classification accuracy. Second, we tackle the speech disorder detection problem as an anomaly detection problem where models are trained merely on typically developing speech, reducing the need for disordered training data. A phoneme-level F 1 score of 0.77 was obtained from an anomaly detection-based system trained on speech attribute features to classify between typical and atypical phoneme pronunciations of children with speech disorder. Finally, we test the efficiency of an x-vector based speaker diarization technique in pediatric therapy sessions. The method successfully distinguished between therapist and child speech with a Diarization Error Rate (DER) of 10%.

In the recent years, extensive researches have been performed on various possible implementations of speaker diarization systems. These systems require efficient clustering algorithms in order to improve their performances in real-time processing. Teaching–learning-based optimization (TLBO) is such clustering algorithm which can be used to resolve the problem to the optimum clustering in a reasonable time. In this paper, a real-time implementation of speaker diarization (SD) system on raspberry pi 3 (RPi 3) using TLBO technique as classifier has been performed. This system has been evaluated on broadcasting radio dataset (NDTV), and the experimental tests have shown that this technique has succeeded to achieve acceptable performances in terms of diarization error rate (DER = 21.90% and 35% in single- and cross-show diarization, respectively), accuracy (87.30%), and real-time factor (RTF = 2.40). Also, we have tested TLBO technique on a 2.4 GHz Intel Core i5 processor using REPERE corpus. Thus, ameliorated results have been obtained in terms of execution time (xRT) and DER in both tasks of single- and cross-show speaker diarization (0.08 and 0.095, and 18.50% and 26.30%, respectively).

Diarization partitions an audio stream into segments based on the voices of the speakers. Real-time diarization systems that include an enrollment step should limit enrollment training samples to reduce user interaction time. Although training on a small number of samples yields poor performance, we show that the accuracy can be improved dramatically using a chronological self-training approach. We studied the tradeoff between training time and classification performance and found that 1 second is sufficient to reach over 95% accuracy. We evaluated on 700 audio conversation files of about 10 minutes each from 6 different languages and demonstrated average diarization error rates as low as 10%., Comment: 5 pages, 5 figures, ICASSP 2021

MEng (Computer Engineering), North-West University, Potchefstroom Campus Speaker diarisation systems aim to answer the question \who spoke when?" and are useful in providing valuable metadata to downstream applications, such as automatic speech recognition systems. However, speaker diarisation systems, like most applications in the field of speech recognition, are especially challenged by domain-mismatch conditions. In this study, we investigate methods with which to adapt a pre-trained diarisation system to a new target domain when only a small in-domain corpus is available and retraining is therefore not an option. We also develop a method for fine-tuning the adaptation process of a pre-trained speaker diarisation system using cluster analysis. Our domain adaptation process focuses on retraining and adapting the statistical components in a speaker diarisation pipeline, which are inherently domain specific, to the target domain. Lastly, we demonstrate this domain adaptation process in a real-world scenario by adapting a pre-trained diarisation system using a small in-domain dataset consisting of telephonic speech from South African call centres. We show that the adapted system can be used to provide metadata which aids the performance of automatic speech recognition systems through speaker-specific adaptations. Masters

Recently developed methods in spontaneous speech analytics require the use of speaker separation based on audio data, referred to as diarization. It is applied to widespread use cases, such as meeting transcription based on recordings from distant microphones and the extraction of the target speaker’s voice profiles from noisy audio. However, speech recognition and analysis can be hindered by background and point-source noise, overlapping speech, and reverberation, which all affect diarization quality in conjunction with each other. To compensate for the impact of these factors, there are a variety of supportive speech analytics methods, such as quality assessments in terms of SNR and RT60 reverberation time metrics, overlapping speech detection, instant speaker number estimation, etc. The improvements in speaker verification methods have benefits in the area of speaker separation as well. This paper introduces several approaches aimed towards improving diarization system quality. The presented experimental results demonstrate the possibility of refining initial speaker labels from neural-based VAD data by means of fusion with labels from quality estimation models, overlapping speech detectors, and speaker number estimation models, which contain CNN and LSTM modules. Such fusing approaches allow us to significantly decrease DER values compared to standalone VAD methods. Cases of ideal VAD labeling are utilized to show the positive impact of ResNet-101 neural networks on diarization quality in comparison with basic x-vectors and ECAPA-TDNN architectures trained on 8 kHz data. Moreover, this paper highlights the advantage of spectral clustering over other clustering methods applied to diarization. The overall quality of diarization is improved at all stages of the pipeline, and the combination of various speech analytics methods makes a significant contribution to the improvement of diarization quality.

The natural language processing community has had a major interest in auto-regressive [4, 13] and span-prediction based language models [7] recently, while knowledge graphs are often referenced for common-sense based reasoning and fact-checking models. In this paper, we present an equivalence representation of span-prediction based language models and knowledge-graphs to better leverage recent developments of language modelling for multi-modal problem statements. Our method performed well, especially with sentiment understanding for multi-modal inputs, and discovered potential bias in naturally occurring videos when compared with movie-data interaction-understanding. We also release a dataset of an auto-generated questionnaire with ground-truths consisting of labels spanning across 120 relationships, 99 sentiments, and 116 interactions, among other labels for finer-grained analysis of model comparisons in the community.

International audience; The demand of high-quality metadata for the available multimedia content requires the development of new techniques able to correctly identify more and more information, including the speaker information. The task known as speaker attribution aims at identifying all or part of the speakers in the audio under analysis. In this work, we carry out a study of the speaker attribution problem in the broadcast domain. Through our experiments, we illustrate the positive impact of diarization on the final performance. Additionally, we show the influence of the variability present in broadcast data, depicting the broadcast domain as a collection of subdomains with particular characteristics. Taking these two factors into account, we also propose alternative approximations robust against domain mismatch. These approximations include a semisupervised alternative as well as a totally unsupervised new hybrid solution fusing diarization and speaker assignment. Thanks to these two approximations, our performance is boosted around a relative 50%. The analysis has been carried out using the corpus for the Albayzín 2020 challenge, a diarization and speaker attribution evaluation working with broadcast data. These data, provided by Radio Televisión Española (RTVE), the Spanish public Radio and TV Corporation, include multiple shows and genres to analyze the impact of new speech technologies in real-world scenarios.

The goal of this paper is to adapt speaker embeddings for solving the problem of speaker diarisation. The quality of speaker embeddings is paramount to the performance of speaker diarisation systems. Despite this, prior works in the field have directly used embeddings designed only to be effective on the speaker verification task. In this paper, we propose three techniques that can be used to better adapt the speaker embeddings for diarisation: dimensionality reduction, attention-based embedding aggregation, and non-speech clustering. A wide range of experiments is performed on various challenging datasets. The results demonstrate that all three techniques contribute positively to the performance of the diarisation system achieving an average relative improvement of 25.07% in terms of diarisation error rate over the baseline., Comment: 5 pages, 2 figures, 3 tables, submitted to Interspeech as a conference paper

Recently, we proposed a novel speaker diarization method called End-to-End-Neural-Diarization-vector clustering (EEND-vector clustering) that integrates clustering-based and end-to-end neural network-based diarization approaches into one framework. The proposed method combines advantages of both frameworks, i.e. high diarization performance and handling of overlapped speech based on EEND, and robust handling of long recordings with an arbitrary number of speakers based on clustering-based approaches. However, the method was only evaluated so far on simulated 2-speaker meeting-like data. This paper is to (1) report recent advances we made to this framework, including newly introduced robust constrained clustering algorithms, and (2) experimentally show that the method can now significantly outperform competitive diarization methods such as Encoder-Decoder Attractor (EDA)-EEND, on CALLHOME data which comprises real conversational speech data including overlapped speech and an arbitrary number of speakers. By further analyzing the experimental results, this paper also discusses pros and cons of the proposed method and reveals potential for further improvement. A set of the code to reproduce the results is available at https://github.com/nttcslab-sp/EEND-vector-clustering., 5 pages, 1 figure, Interspeech2021. (Update to include a reference to the code)

Speaker diarization accuracy can be affected by both acoustics and conversation characteristics. Determining the cause of diarization errors is difficult because speaker voice acoustics and conversation structure co-vary, and the interactions between acoustics, conversational structure, and diarization accuracy are complex. This paper proposes a methodology that can distinguish independent marginal effects of acoustic and conversation characteristics on diarization accuracy by remixing conversations in a factorial design. As an illustration, this approach is used to investigate gender-related and language-related accuracy differences with three diarization systems: a baseline system using subsegment x-vector clustering, a variant of it with shorter subsegments, and a third system based on a Bayesian hidden Markov model. Our analysis shows large accuracy disparities for the baseline system primarily due to conversational structure, which are partially mitigated in the other two systems. The illustration thus demonstrates how the methodology can be used to identify and guide diarization model improvements., 5 pages, 2 figures, Interspeech 2021

In this work, we present a novel audio-visual dataset for active speaker detection in the wild. A speaker is considered active when his or her face is visible and the voice is audible simultaneously. Although active speaker detection is a crucial pre-processing step for many audio-visual tasks, there is no existing dataset of natural human speech to evaluate the performance of active speaker detection. We therefore curate the Active Speakers in the Wild (ASW) dataset which contains videos and co-occurring speech segments with dense speech activity labels. Videos and timestamps of audible segments are parsed and adopted from VoxConverse, an existing speaker diarisation dataset that consists of videos in the wild. Face tracks are extracted from the videos and active segments are annotated based on the timestamps of VoxConverse in a semi-automatic way. Two reference systems, a self-supervised system and a fully supervised one, are evaluated on the dataset to provide the baseline performances of ASW. Cross-domain evaluation is conducted in order to show the negative effect of dubbed videos in the training data.

In this paper, we describe novel components for extracting clinically relevant information from medical conversations which will be available as Google APIs. We describe a transformer-based Recurrent Neural Network Transducer (RNN-T) model tailored for long-form audio, which can produce rich transcriptions including speaker segmentation, speaker role labeling, punctuation and capitalization. On a representative test set, we compare performance of RNN-T models with different encoders, units and streaming constraints. Our transformer-based streaming model performs at about 20% WER on the ASR task, 6% WDER on the diarization task, 43% SER on periods, 52% SER on commas, 43% SER on question marks and 30% SER on capitalization. Our recognizer is paired with a confidence model that utilizes both acoustic and lexical features from the recognizer. The model performs at about 0.37 NCE. Finally, we describe a RNN-T based tagging model. The performance of the model depends on the ontologies, with F-scores of 0.90 for medications, 0.76 for symptoms, 0.75 for conditions, 0.76 for diagnosis, and 0.61 for treatments. While there is still room for improvement, our results suggest that these models are sufficiently accurate for practical applications.

We recently proposed DOVER-Lap, a method for combining overlap-aware speaker diarization system outputs. DOVER-Lap improved upon its predecessor DOVER by using a label mapping method based on globally-informed greedy search. In this paper, we analyze this label mapping in the framework of a maximum orthogonal graph partitioning problem, and present three inferences. First, we show that DOVER-Lap label mapping is exponential in the input size, which poses a challenge when combining a large number of hypotheses. We then revisit the DOVER label mapping algorithm and propose a modification which performs similar to DOVER-Lap while being computationally tractable. We also derive an approximation bound for the algorithm in terms of the maximum number of hypotheses speakers. Finally, we describe a randomized local search algorithm which provides a near-optimal $(1-\epsilon)$-approximate solution to the problem with high probability. We empirically demonstrate the effectiveness of our methods on the AMI meeting corpus. Our code is publicly available: https://github.com/desh2608/dover-lap., Comment: 5 pages, 3 figures. Acceped at INTERSPEECH 2021

Target-speaker voice activity detection (TS-VAD) has recently shown promising results for speaker diarization on highly overlapped speech. However, the original model requires a fixed (and known) number of speakers, which limits its application to real conversations. In this paper, we extend TS-VAD to speaker diarization with unknown numbers of speakers. This is achieved by two steps: first, an initial diarization system is applied for speaker number estimation, followed by TS-VAD network output masking according to this estimate. We further investigate different diarization methods, including clustering-based and region proposal networks, for estimating the initial i-vectors. Since these systems have complementary strengths, we propose a fusion-based method to combine frame-level decisions from the systems for an improved initialization. We demonstrate through experiments on variants of the LibriCSS meeting corpus that our proposed approach can improve the DER by up to 50\% relative across varying numbers of speakers. This improvement also results in better downstream ASR performance approaching that using oracle segments.

In our previous work, we introduced our Bayesian Hidden Markov Model with eigenvoice priors, which has been recently recognized as the state-of-the-art model for Speaker Diarization. In this article we present a more complete analysis of the Diarization system. The inference of the model is fully described and derivations of all update formulas are provided for a complete understanding of the algorithm. An extensive analysis on the effect, sensitivity and interactions of all model parameters is provided, which might be used as a guide for their optimal setting. The newly introduced speaker regularization coefficient allows us to control the number of speakers inferred in an utterance. A naive speaker model merging strategy is also presented, which allows to drive the variational inference out of local optima. Experiments for the different diarization scenarios are presented on CALLHOME and DIHARD datasets.

Speaker diarization system identifies the speaker homogenous regions in those set of recordings where multiple speakers are present. It answers the question ‘who spoke when?’. The data set for speaker diarization usually consists of telephone, meetings, TV/ talk shows, broadcast news and other multi-speaker recordings. In this paper, we present the performance of our proposed multimodal speaker diarization system under noisy conditions. Two types of noises comprising additive white Gaussian noise (AWGN) and realistic environmental noise is used to evaluate the system. To mitigate the effect of noise, we propose to add an LSTM based speech enhancement block in our diarization pipeline. This block is trained on synthesized data set with more than 100 noise types to enhance the noisy speech. The enhanced speech is further used in multimodal speaker diarization system which utilizes a pre-trained audio-visual synchronization model to find the active speaker. High confidence active speaker segments are then used to train the speaker specific clusters on the enhanced speech. A subset of AMI corpus consisting of 5.4 h of recordings is used in this analysis. For AWGN, the LSTM model performance improvement is comparable with Wiener filter while in case of realistic environmental noise, the LSTM model improves significantly as compared to Wiener filter in terms of diarization error rate (DER).

In this paper, we address the problem of optimal non-hierarchical clustering in the speaker clustering phase for the speaker diarization task of news broadcasts. A new hybridization combining differential evolution (DE) algorithm and K-means algorithm is proposed and tested on TV news database (TVND). To optimize the classification of speakers, two criteria, namely trace within criterion (TRW) and variance ratio criterion (VRC), were used as clustering validity indices, correcting every possible grouping of speakers’ segments. Concerning the encoding of the classification of clusters to be optimized, it is performed by the cluster centers in DE algorithm. Therefore, a problem of rearrangement of centers in the populations can be generated, which cannot ensure an efficient search by applying evolutionary operators. For this purpose, an efficient heuristic was also proposed for this rearrangement. Non-hybrid DE variants were applied with and without the rearrangement of cluster centers, and compared with the corresponding hybrid K-means variants. The experimental results have showed the high-efficiency of hybrid K-means variants with the rearrangement of cluster centers compared with those without the rearrangement of cluster centers and non-hybrid DE variants. Also, the obtained results using hybrid and non-hybrid DE variants with the rearrangement of cluster centers were quite similar using both TWR and VRC criteria. Moreover, the best efficiency was acquired using hybrid DE variants thanks to these two criteria from which a value of 13.05% of DER has been reached by hybrid b6e6rl variant.

This paper deals with the problem of Voice Activity Detection (VAD) used in speaker diarization and detection systems to detect silent areas in speech stream. In fact, an incorrect VAD can cause serious performances degradation of these systems. To overcome this problem, we propose in this work, a new VAD architecture which combines two VAD. The first one is employed in the temporal domain and the second one is performed in the cepstral domain. Classical VAD architectures used either temporal VAD or cepstral one to eliminate the silent segments in audio documents. In this work, we propose to use the both VAD in order to reduce silent detection errors. This new VAD architecture is evaluated on speaker diarization and detection systems using the BIC distance (Bayesian Information Criterion), two different classifiers, Gaussian Mixture Model (GMM) classifier for the segmentation task and Hierarchical Ascending Clustering (HAC) or Support Vector Machine (SVM) classifier for the clustering task. A comparative investigation was done between our new proposed VAD architecture and the classical ones applied in the preprocessing step of the speaker diarization/detection systems. For the evaluation task, the different VAD architectures were tested on telephonic conversations extracted from the NIST-2005 corpus. The results of experiments have shown that the new VAD architecture has considerably enhanced the performances of speaker diarization/detection systems, comparing to the classical VAD architectures. The results obtained by the new VAD architecture are illustrated by a Diarization Error Rate (DER) of only 2.78% and a Speaker Detection Rate (SDR) of 97.67%.

We propose a novel speaker-dependent speech separation framework for the challenging CHiME-5 acoustic environments, exploiting advantages of both deep learning based and conventional preprocessing techniques to prepare data effectively for separating target speech from multi-talker mixed speech collected with multiple microphone arrays. First, a series of multi-channel operations is conducted to reduce existing reverberation and noise, and a single-channel deep learning based speech enhancement model is used to predict speech presence probabilities. Next, a two-stage supervised speech separation approach, using oracle speaker diarization information from CHiME-5, is proposed to separate speech of a target speaker from interference speakers in mixed speech. Given a set of three estimated masks of the background noise, the target speaker and the interference speakers from single-channel speech enhancement and separation models, a complex Gaussian mixture model based generalized eigenvalue beamformer is then used for enhancing the signal at the reference array while avoiding the speaker permutation issue. Furthermore, the proposed front-end can generate a large variety of processed data for an ensemble of speech recognition results. Experiments on the development set have shown that the proposed two-stage approach can yield significant improvements of recognition performance over the official baseline system and achieved top accuracies in all four competing evaluation categories among all systems submitted to the CHiME-5 Challenge.

Understanding and assessing child verbal communication patterns is critical in facilitating effective language development. Typically speaker diarization is performed to explore children’s verbal engagement. Understanding which activity areas stimulate verbal communication can help promote more efficient language development. In this study, we present a two-stage children vocal engagement prediction system that consists of (1) a near to real-time, noise robust system that measures the duration of child-to-adult and child-to-child conversations, and tracks the number of conversational turn-takings, (2) a novel child location tracking strategy, that determines in which activity areas a child spends most/least of their time. A proposed child–adult turn-taking solution relies exclusively on vocal cues observed during the interaction between a child and other children, and/or classroom teachers. By employing a threshold optimized speech activity detection using a linear combination of voicing measures, it is possible to achieve effective speech/non-speech segment detection prior to conversion assessment. This TO-COMBO-SAD reduces classification error rates for adult-child audio by 21.34% and 27.3% compared to a baseline i-Vector and standard Bayesian Information Criterion diarization systems, respectively. In addition, this study presents a unique location tracking system adult-child that helps determine the quantity of child–adult communication in specific activity areas, and which activities stimulate voice communication engagement in a child–adult education space. We observe that our proposed location tracking solution offers unique opportunities to assess speech and language interaction for children, and quantify the location context which would contribute to improve verbal communication.

Spoken language diarization (SLD) is a task to perform automatic segmentation and labeling of the languages present in a given code-switched speech utterance. Inspiring from the way humans perform SLD (i.e capturing the language specific long term information), this work has proposed an acoustic-phonetic approach to perform SLD. This acoustic phonetic approach consists of an attention based neural network modelling to capture the language specific information and a Gaussian smoothing approach to locate the language change points. From the experimental study, it has been observed that the proposed approach performs better when dealing with code-switched segment containing monolingual segments of longer duration. However, the performance of the approach decreases with decrease in the monolingual segment duration. This issue poses a challenge in the further exploration of the proposed approach.

In this paper we describe a speaker diarization system that enables localization and identification of all speakers present in a conversation or meeting. We propose a novel systematic approach to tackle several long-standing challenges in speaker diarization tasks: (1) to segment and separate overlapping speech from two speakers; (2) to estimate the number of speakers when participants may enter or leave the conversation at any time; (3) to provide accurate speaker identification on short text-independent utterances; (4) to track down speakers movement during the conversation; (5) to detect speaker change incidence real-time. First, a differential directional microphone array-based approach is exploited to capture the target speakers' voice in far-field adverse environment. Second, an online speaker-location joint clustering approach is proposed to keep track of speaker location. Third, an instant speaker number detector is developed to trigger the mechanism that separates overlapped speech. The results suggest that our system effectively incorporates spatial information and achieves significant gains., Published in ICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)

In this paper, we propose a classroom management system in which the attendance of the students will be recorded using face recognition. Moreover, the system will also help the teacher to identify the students who are talking during the lecture by speech diarization algorithm. The proposed system will essentially help the teacher to conduct the lecture smoothly without actually investing any time in taking attendance or identifying the students who are creating a nuisance.

The technique used for recognizing a language by utilizing pronounced speech is called spoken Language Identification (LID). This field has a high significance in the interaction between human and computer. Besides, it can be implemented in several applications such as call centers, speaker diarization in multilingual environments, and in translation systems using a speech-to-speech manner. However, most studies that used LID systems are used and focused on neutral speech only. Moreover, the application of emotional speech in LID systems is crucial in real applications. Therefore, this study aims to investigate the performance of Extreme Learning Machine (ELM) in LID system by utilizing emotional speech. The system is evaluated based on two different languages (Germany and English language). This study has used the Berlin Emotional Speech Dataset (BESD) for the Germany language while the Ryerson Audio-Visual Dataset of Emotional Speech and Song (RAVDESS) for the English language. Four different evaluation scenarios (All Dataset (AD), Normal-Speech Dependent (N-SD), Gender-Female Dependent (G-FD), and Gender-Male Dependent (G-MD) scenario) have been conducted in order to evaluate the system. The experiments results have shown that the highest performance was achieved an accuracy of 99.08%, 100.00%, 98.22%, and 99.37% for AD, N-SD, G-FD, and G-MD scenario, respectively.

In this paper, we present a semi-supervised training technique using pseudo-labeling for end-to-end neural diarization (EEND). The EEND system has shown promising performance compared with traditional clustering-based methods, especially in the case of overlapping speech. However, to get a well-tuned model, EEND requires labeled data for all the joint speech activities of every speaker at each time frame in a recording. In this paper, we explore a pseudo-labeling approach that employs unlabeled data. First, we propose an iterative pseudo-label method for EEND, which trains the model using unlabeled data of a target condition. Then, we also propose a committee-based training method to improve the performance of EEND. To evaluate our proposed method, we conduct the experiments of model adaptation using labeled and unlabeled data. Experimental results on the CALLHOME dataset show that our proposed pseudo-label achieved a 37.4% relative diarization error rate reduction compared to a seed model. Moreover, we analyzed the results of semi-supervised adaptation with pseudo-labeling. We also show the effectiveness of our approach on the third DIHARD dataset., Accepted for Interspeech 2021

Identifying the identity of the speaker of short segments in human dialogue has been considered one of the most challenging problems in speech signal processing. Speaker representations of short speech segments tend to be unreliable, resulting in poor fidelity of speaker representations in tasks requiring speaker recognition. In this paper, we propose an unconventional method that tackles the trade-off between temporal resolution and the quality of the speaker representations. To find a set of weights that balance the scores from multiple temporal scales of segments, a neural affinity score fusion model is presented. Using the CALLHOME dataset, we show that our proposed multi-scale segmentation and integration approach can achieve a state-of-the-art diarization performance., Submitted to ICASSP 2021

This paper investigates the utilization of an end-to-end diarization model as post-processing of conventional clustering-based diarization. Clustering-based diarization methods partition frames into clusters of the number of speakers; thus, they typically cannot handle overlapping speech because each frame is assigned to one speaker. On the other hand, some end-to-end diarization methods can handle overlapping speech by treating the problem as multi-label classification. Although some methods can treat a flexible number of speakers, they do not perform well when the number of speakers is large. To compensate for each other's weakness, we propose to use a two-speaker end-to-end diarization method as post-processing of the results obtained by a clustering-based method. We iteratively select two speakers from the results and update the results of the two speakers to improve the overlapped region. Experimental results show that the proposed algorithm consistently improved the performance of the state-of-the-art methods across CALLHOME, AMI, and DIHARD II datasets.

We present a novel online end-to-end neural diarization system, BW-EDA-EEND, that processes data incrementally for a variable number of speakers. The system is based on the Encoder-Decoder-Attractor (EDA) architecture of Horiguchi et al., but utilizes the incremental Transformer encoder, attending only to its left contexts and using block-level recurrence in the hidden states to carry information from block to block, making the algorithm complexity linear in time. We propose two variants: For unlimited-latency BW-EDA-EEND, which processes inputs in linear time, we show only moderate degradation for up to two speakers using a context size of 10 seconds compared to offline EDA-EEND. With more than two speakers, the accuracy gap between online and offline grows, but the algorithm still outperforms a baseline offline clustering diarization system for one to four speakers with unlimited context size, and shows comparable accuracy with context size of 10 seconds. For limited-latency BW-EDA-EEND, which produces diarization outputs block-by-block as audio arrives, we show accuracy comparable to the offline clustering-based system.

Estimating the positions of multiple speakers can be helpful for tasks like automatic speech recognition or speaker diarization. Both applications benefit from a known speaker position when, for instance, applying beamforming or assigning unique speaker identities. Recently, several approaches utilizing acoustic signals augmented with visual data have been proposed for this task. However, both the acoustic and the visual modality may be corrupted in specific spatial regions, for instance due to poor lighting conditions or to the presence of background noise. This paper proposes a novel audiovisual data fusion framework for speaker localization by assigning individual dynamic stream weights to specific regions in the localization space. This fusion is achieved via a neural network, which combines the predictions of individual audio and video trackers based on their time- and location-dependent reliability. A performance evaluation using audiovisual recordings yields promising results, with the proposed fusion approach outperforming all baseline models., 4 pages, 6 figures, ICASSP 2021

Target speech extraction, which extracts the speech of a target speaker in a mixture given auxiliary speaker clues, has recently received increased interest. Various clues have been investigated such as pre-recorded enrollment utterances, direction information, or video of the target speaker. In this paper, we explore the use of speaker activity information as an auxiliary clue for single-channel neural network-based speech extraction. We propose a speaker activity driven speech extraction neural network (ADEnet) and show that it can achieve performance levels competitive with enrollment-based approaches, without the need for pre-recordings. We further demonstrate the potential of the proposed approach for processing meeting-like recordings, where the speaker activity is obtained from a diarization system. We show that this simple yet practical approach can successfully extract speakers after diarization, which results in improved ASR performance, especially in high overlapping conditions, with a relative word error rate reduction of up to 25%., Comment: To appear in ICASSP 2021