skip to main content

This content will become publicly available on September 18, 2023

Title: Improving Speech Emotion Recognition Using Self-Supervised Learning with Domain-Specific Audiovisual Tasks
Speech emotion recognition (SER) is a challenging task due to the limited availability of real-world labeled datasets. Since it is easier to find unlabeled data, the use of self-supervised learning (SSL) has become an attractive alternative. This study proposes new pre-text tasks for SSL to improve SER. While our target application is SER, the proposed pre-text tasks include audio-visual formulations, leveraging the relationship between acoustic and facial features. Our proposed approach introduces three new unimodal and multimodal pre-text tasks that are carefully designed to learn better representations for predicting emotional cues from speech. Task 1 predicts energy variations (high or low) from a speech sequence. Task 2 uses speech features to predict facial activation (high or low) based on facial landmark movements. Task 3 performs a multi-class emotion recognition task on emotional labels obtained from combinations of action units (AUs) detected across a video sequence. We pre-train a network with 60.92 hours of unlabeled data, fine-tuning the model for the downstream SER task. The results on the CREMA-D dataset show that the model pre-trained on the proposed domain-specific pre-text tasks significantly improves the precision (up to 5.1%), recall (up to 4.5%), and F1-scores (up to 4.9%) of our SER system.
Award ID(s):
Publication Date:
Journal Name:
Interspeech 2022
Page Range or eLocation-ID:
1168 to 1172
Sponsoring Org:
National Science Foundation
More Like this
  1. Semi-supervised learning (SSL) is an appealing approach to resolve generalization problem for speech emotion recognition (SER) systems. By utilizing large amounts of unlabeled data, SSL is able to gain extra information about the prior distribution of the data. Typically, it can lead to better and robust recognition performance. Existing SSL approaches for SER include variations of encoder-decoder model structures such as autoencoder (AE) and variational autoencoders (VAEs), where it is difficult to interpret the learning mechanism behind the latent space. In this study, we introduce a new SSL framework, which we refer to as the DeepEmoCluster framework, for attribute-based SER tasks. The DeepEmoCluster framework is an end-to-end model with mel-spectrogram inputs, which combines a self-supervised pseudo labeling classification network with a supervised emotional attribute regressor. The approach encourages the model to learn latent representations by maximizing the emotional separation of K-means clusters. Our experimental results based on the MSP-Podcast corpus indicate that the DeepEmoCluster framework achieves competitive prediction performances in fully supervised scheme, outperforming baseline methods in most of the conditions. The approach can be further improved by incorporating extra unlabeled set. Moreover, our experimental results explicitly show that the latent clusters have emotional dependencies, enriching the geometric interpretation ofmore »the clusters.« less
  2. Speech emotion recognition (SER) plays an important role in multiple fields such as healthcare, human-computer interaction (HCI), and security and defense. Emotional labels are often annotated at the sentence-level (i.e., one label per sentence), resulting in a sequence-to-one recognition problem. Traditionally, studies have relied on statistical descriptions, which are com- puted over time from low level descriptors (LLDs), creating a fixed dimension sentence-level feature representation regardless of the duration of the sentence. However sentence-level features lack temporal information, which limits the performance of SER systems. Recently, new deep learning architectures have been proposed to model temporal data. An important question is how to extract emotion-relevant features with temporal infor- mation. This study proposes a novel data processing approach that extracts a fixed number of small chunks over sentences of different durations by changing the overlap between these chunks. The approach is flexible, providing an ideal frame- work to combine gated network or attention mechanisms with long short-term memory (LSTM) networks. Our experimental results based on the MSP-Podcast dataset demonstrate that the proposed method not only significantly improves recognition accuracy over alternative temporal-based models relying on LSTM, but also leads to computational efficiency.
  3. A critical issue of current speech-based sequence-to-one learning tasks, such as speech emotion recognition (SER), is the dynamic temporal modeling for speech sentences with different durations. The goal is to extract an informative representation vector of the sentence from acoustic feature sequences with varied length. Traditional methods rely on static descriptions such as statistical functions or a universal background model (UBM), which are not capable of characterizing dynamic temporal changes. Recent advances in deep learning architectures provide promising results, directly extracting sentence-level representations from frame-level features. However, conventional cropping and padding techniques that deal with varied length sequences are not optimal, since they truncate or artificially add sentence-level information. Therefore, we propose a novel dynamic chunking approach, which maps the original sequences of different lengths into a fixed number of chunks that have the same duration by adjusting their overlap. This simple chunking procedure creates a flexible framework that can incorporate different feature extractions and sentence-level temporal aggregation approaches to cope, in a principled way, with different sequence-to-one tasks. Our experimental results based on three databases demonstrate that the proposed framework provides: 1) improvement in recognition accuracy, 2) robustness toward different temporal length predictions, and 3) high model computational efficiencymore »advantages.« less
  4. The orofacial area conveys a range of information, including speech articulation and emotions. These two factors add constraints to the facial movements, creating non-trivial integrations and interplays. To generate more expressive and naturalistic movements for conversational agents (CAs) the relationship between these factors should be carefully modeled. Data-driven models are more appropriate for this task than rule-based systems. This paper provides two deep learning speech-driven structures to integrate speech articulation and emotional cues. The proposed approaches rely on multitask learning (MTL) strategies, where related secondary tasks are jointly solved when synthesizing orofacial movements. In particular, we evaluate emotion recognition and viseme recognition as secondary tasks. The approach creates shared representations that generate behaviors that not only are closer to the original orofacial movements, but also are perceived more natural than the results from single task learning.
  5. A challenging task in affective computing is to build reliable speech emotion recognition (SER) systems that can accurately predict emotional attributes from spontaneous speech. To increase the trust in these SER systems, it is important to predict not only their accuracy, but also their confidence. An intriguing approach to predict uncertainty is Monte Carlo (MC) dropout, which obtains pre- dictions from multiple feed-forward passes through a deep neural network (DNN) by using dropout regularization in both training and inference. This study evaluates this approach with regression models to predict emotional attribute scores for valence, arousal and dom- inance. The analysis illustrates that predicting uncertainty in this problem is possible, where the performance is higher for samples in the test set with lower uncertainty. The study evaluates uncertainty estimation as a function of the emotional attributes, showing that samples with extreme values have lower uncertainty. Finally, we demonstrate the benefits of uncertainty estimation with reject option, where a classifier can decline to give a prediction when its confi- dence is low. By rejecting only 25% of the test set with the highest uncertainty, we achieve relative performance gains of 7.34% for arousal, 13.73% for valence and 8.79% for dominance.