Search

Speech emotion recognition (SER) systems find applications in various fields such as healthcare, education, and security and defense. A major drawback of these systems is their lack of generalization across different conditions. For example, systems that show superior performance on certain databases show poor performance when tested on other corpora. This problem can be solved by training models on large amounts of labeled data from the target domain, which is expensive and time-consuming. Another approach is to increase the generalization of the models. An effective way to achieve this goal is by regularizing the models through multitask learning (MTL), where auxiliary tasks are learned along with the primary task. These methods often require the use of labeled data which is computationally expensive to collect for emotion recognition (gender, speaker identity, age or other emotional descriptors). This study proposes the use of ladder networks for emotion recognition, which utilizes an unsupervised auxiliary task. The primary task is a regression problem to predict emotional attributes. The auxiliary task is the reconstruction of intermediate feature representations using a denoising autoencoder. This auxiliary task does not require labels so it is possible to train the framework in a semi-supervised fashion with abundant unlabeled data from the target domain. This study shows that the proposed approach creates a powerful framework for SER, achieving superior performance than fully supervised single-task learning (STL) and MTL baselines. We implement the approach with sentence-level or frame-level features, demonstrating the flexibility of our approach. Additionally, the generalization of the ladder networks is evaluated in cross-corpus settings using sentence-level features, obtaining important improvements. Compared to the STL baselines, the proposed approach achieves relative gains in concordance correlation coefficient (CCC) between 3.0% and 3.5% for within corpus evaluations, and between 16.1% and 74.1% for cross corpus evaluations, highlighting the power of the architecture.

General embeddings like word2vec, GloVe and ELMo have shown a lot of success in natural language tasks. The embeddings are typically extracted from models that are built on general tasks such as skip-gram models and natural language generation. In this paper, we extend the work from natural language understanding to multi-modal architectures that use audio, visual and textual information for machine learning tasks. The embeddings in our network are extracted using the encoder of a transformer model trained using multi-task training. We use person identification and automatic speech recognition as the tasks in our embedding generation framework. We tune and evaluate the embeddings on the downstream task of emotion recognition and demonstrate that on the CMU-MOSEI dataset, the embeddings can be used to improve over previous state of the art results., To appear in Interspeech,2020

This paper presents an audio visual automatic speech recognition (AV-ASR) system using a Transformer-based architecture. We particularly focus on the scene context provided by the visual information, to ground the ASR. We extract representations for audio features in the encoder layers of the transformer and fuse video features using an additional crossmodal multihead attention layer. Additionally, we incorporate a multitask training criterion for multiresolution ASR, where we train the model to generate both character and subword level transcriptions. Experimental results on the How2 dataset, indicate that multiresolution training can speed up convergence by around 50% and relatively improves word error rate (WER) performance by upto 18% over subword prediction models. Further, incorporating visual information improves performance with relative gains upto 3.76% over audio only models. Our results are comparable to state-of-the-art Listen, Attend and Spell-based architectures.

Conventional models for emotion recognition from speech signal are trained in supervised fashion using speech utterances with emotion labels. In this study we hypothesize that speech signal depends on multiple latent variables including the emotional state, age, gender, and speech content. We propose an Adversarial Autoencoder (AAE) to perform variational inference over the latent variables and reconstruct the input feature representations. Reconstruction of feature representations is used as an auxiliary task to aid the primary emotion recognition task. Experiments on the IEMOCAP dataset demonstrate that the auxiliary learning tasks improve emotion classification accuracy compared to a baseline supervised classifier. Further, we demonstrate that the proposed learning approach can be used for the end-to-end speech emotion recognition, as its applicable for models that operate on frame-level inputs.

We present the MSP-IMPROV corpus, a multimodal emotional database, where the goal is to have control over lexical content and emotion while also promoting naturalness in the recordings. Studies on emotion perception often require stimuli with fixed lexical content, but that convey different emotions. These stimuli can also serve as an instrument to understand how emotion modulates speech at the phoneme level, in a manner that controls for coarticulation. Such audiovisual data are not easily available from natural recordings. A common solution is to record actors reading sentences that portray different emotions, which may not produce natural behaviors. We propose an alternative approach in which we define hypothetical scenarios for each sentence that are carefully designed to elicit a particular emotion. Two actors improvise these emotion-specific situations, leading them to utter contextualized, non-read renditions of sentences that have fixed lexical content and convey different emotions. We describe the context in which this corpus was recorded, the key features of the corpus, the areas in which this corpus can be useful, and the emotional content of the recordings. The paper also provides the performance for speech and facial emotion classifiers. The analysis brings novel classification evaluations where we study the performance in terms of inter-evaluator agreement and naturalness perception, leveraging the large size of the audiovisual database.

Manual annotations and transcriptions have an ever-increasing importance in areas such as behavioral signal processing, image processing, computer vision, and speech signal processing. Conventionally, this metadata has been collected through manual annotations by experts. With the advent of crowdsourcing services, the scientific community has begun to crowdsource many tasks that researchers deem tedious, but can be easily completed by many human annotators. While crowdsourcing is a cheaper and more efficient approach, the quality of the annotations becomes a limitation in many cases. This paper investigates the use of reference sets with predetermined ground-truth to monitor annotators’ accuracy and fatigue, all in real-time. The reference set includes evaluations that are identical in form to the relevant questions that are collected, so annotators are blind to whether or not they are being graded on performance on a specific question. We explore these ideas on the emotional annotation of the MSP-IMPROV database. We present promising results which suggest that our system is suitable for collecting accurate annotations.

Recognizing emotions using few attribute dimensions such as arousal, valence and dominance provides the flexibility to effectively represent complex range of emotional behaviors. Conventional methods to learn these emotional descriptors primarily focus on separate models to recognize each of these attributes. Recent work has shown that learning these attributes together regularizes the models, leading to better feature representations. This study explores new forms of regularization by adding unsupervised auxiliary tasks to reconstruct hidden layer representations. This auxiliary task requires the denoising of hidden representations at every layer of an auto-encoder. The framework relies on ladder networks that utilize skip connections between encoder and decoder layers to learn powerful representations of emotional dimensions. The results show that ladder networks improve the performance of the system compared to baselines that individually learn each attribute, and conventional denoising autoencoders. Furthermore, the unsupervised auxiliary tasks have promising potential to be used in a semi-supervised setting, where few labeled sentences are available., Comment: Submitted to Interspeech 2018

Affect recognition plays an important role in human computer interaction (HCI). Speech is one of the primary forms of expression and is an important modality for emotion recognition. While multiple recognition systems exist, the most common ones identify discrete categories such as happiness, sadness, from distinct utterances that are a few seconds long. In many cases the datasets, used for training and evaluation, are imbalanced across the emotion labels. This leads to big discrepancies between the unweighted accuracy (UA) and weighted accuracy (WA). Recently Deep Neural Networks have shown increased performance for the emotion classification task. In particular Convolutional Neural Networks capture contextual information from speech feature frames. In this paper we analyze various convolutional architectures for speech emotion recognition. We report performance on different frame level features. Further we analyze various pooling techniques, on top of convolutional layers, to get a utterance level representation for the emotion. Our best system provides a performance of UA + WA of 121.15 compared to the baseline algorithm performance of 118.10.

Studies have shown that ranking emotional attributes through preference learning methods has significant advantages over conventional emotional classification/regression frameworks. Preference learning is particularly appealing for retrieval tasks, where the goal is to identify speech conveying target emotional behaviors (e.g., positive samples with low arousal). With recent advances in deep neural networks (DNNs), this study explores whether a preference learning framework relying on deep learning can outperform conventional ranking algorithms. We use a deep learning ranker implemented with the RankNet algorithm to evaluate preference between emotional sentences in terms of dimensional attributes (arousal, valence and dominance). The results show improved performance over ranking algorithms trained with support vector machine (SVM) (i.e., RankSVM). The results are significantly better than performance reported in previous work, demonstrating the potential of RankNet to retrieve speech with target emotional behaviors.

Expressive speech introduces variations in the acoustic features affecting the performance of speech technology such as speaker verification systems. It is important to identify the range of emotions for which we can reliably estimate speaker verification tasks. This paper studies the performance of a speaker verification system as a function of emotions. Instead of categorical classes such as happiness or anger, which have important intra-class variability, we use the continuous attributes arousal, valence, and dominance which facilitate the analysis. We evaluate an speaker verification system trained with the i-vector framework with a probabilistic linear discriminant analysis (PLDA) back-end. The study relies on a subset of the MSP-PODCAST corpus, which has naturalistic recordings from 40 speakers. We train the system with neutral speech, creating mismatches on the testing set. The results show that speaker verification errors increase when the values of the emotional attributes increase. For neutral/moderate values of arousal, valence and dominance, the speaker verification performance are reliable. These results are also observed when we artificially force the sentences to have the same duration.

Research on automatic speech summarization typically focuses on optimizing objective evaluation criteria, such as the ROUGE metric, which depend on word and phrase overlaps between automatic and manually generated summary documents. However, the actual quality of the speech summarizer largely depends on how the end-users perceive the audio output. This work focuses on the task of composing summarized audio streams with the aim of improving the quality and interest perceived by the end-user. First, using crowd-sourced summary annotations on a broadcast news corpus, we train a rank-SVM classifier to learn the relative importance of each sentence in a news story. Acoustic, lexical and structural features are used for training. In addition, we investigate the perceived emotion level in each sentence to aid the summarizer in selecting interesting sentences, yielding an emotion-aware summarizer. Next, we propose several methods to combine these sentences to generate a compressed audio stream. Subjective evaluations are performed to evaluate the quality of the generated summaries on the following criterion: interest, abruptness, informativeness, attractiveness, and overall quality. The results indicate that users are most sensitive to the linguistic coherence and continuity of the audio stream.