An official website of the United States government
Training acoustic models with sequentially incoming data – while both leveraging new data and avoiding the forgetting effect – is an essential obstacle to achieving human intelligence level in speech recognition. An obvious approach to leverage data from a new domain (e.g., new accented speech) is to first generate a comprehensive dataset of all domains, by combining all available data, and then use this dataset to retrain the acoustic models. However, as the amount of training data grows, storing and retraining on such a large-scale dataset becomes practically impossible. To deal with this problem, in this study, we study several domain expansion techniques which exploit only the data of the new domain to build a stronger model for all domains. These techniques are aimed at learning the new domain with a minimal forgetting effect (i.e., they maintain original model performance). These techniques modify the adaptation procedure by imposing new constraints including (1) weight constraint adaptation (WCA): keeping the model parameters close to the original model parameters; (2) elastic weight consolidation (EWC): slowing down training for parameters that are important for previously established domains; (3) soft KL-divergence (SKLD): restricting the KL-divergence between the original and the adapted model output distributions; and (4) hybrid SKLD-EWC: incorporating both SKLD and EWC constraints. We evaluate these techniques in an accent adaptation task in which we adapt a deep neural network (DNN) acoustic model trained with native English to three different English accents: Australian, Hispanic, and Indian. The experimental results show that SKLD significantly outperforms EWC, and EWC works better than WCA. The hybrid SKLD-EWC technique results in the best overall performance.
more »
« less
Domain Expansion for End-to-End Speech Recognition: Applications for Accent/Dialect Speech
Training Automatic Speech Recognition (ASR) systems with sequentially incoming data from alternate domains is an essential milestone in order to reach human intelligibility level in speech recognition. The main challenge of sequential learning is that current adaptation techniques result in significant performance degradation for previously-seen domains.To mitigate the catastrophic forgetting problem, this study proposes effective domain expansion techniques for two scenarios: 1)where only new domain data is available, and 2) where both prior and new domain data are available. We examine the efficacy of the approaches through experiments on adapting a model trained with native English to different English accents. For the first scenario, we study several existing and proposed regularization-based approaches to mitigate performance loss of initial data.The experiments demonstrate the superior performanceo four proposed Soft KL-Divergence(SKLD)-Model Averaging (MA) approach. In this approach, SKLD first alleviates the forgetting problem during adaptation; next, MA makes the final efficient compromise between the two domains by averaging parameters of the initial and adapted models. For the second scenario, we explore several rehearsal-based approaches, which leverage initial data to maintain the original model performance.We propose Gradient Averaging (GA) as well as an approach which operates by averaging gradients computed for both initial and new domains. Experiments demonstrate that GA outperforms retraining and specifically designed continual learning approaches, such as Averaged Gradient Episodic Memory (AGEM). Moreover, GA significantly improves computational costs over the complete retraining approach.
more »
« less
- Award ID(s):
- 1918032
- PAR ID:
- 10478761
- Publisher / Repository:
- IEEE
- Date Published:
- Journal Name:
- IEEE/ACM Transactions on Audio, Speech, and Language Processing
- Volume:
- 31
- ISSN:
- 2329-9290
- Page Range / eLocation ID:
- 762 - 774
- Subject(s) / Keyword(s):
- Accented speech continuallearning domain expansion end-to-end systems model adaptation speech recognition
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Recently, speech foundation models have gained popularity due to their superiority in finetuning downstream ASR tasks. However, models finetuned on certain domains, such as LibriSpeech (adult read speech), behave poorly on other domains (child or noisy speech). One solution could be collecting as much labeled and diverse data as possible for joint finetuning on various domains. However, collecting target domain speech-text paired data and retraining the model is often costly and computationally expensive. In this paper, we introduce a simple yet effective method, speech only adaptation (SOA), based on speech foundation models (Wav2vec 2.0), which requires only speech input data from the target domain. Specifically, the Wav2vec 2.0 feature encoder is continually pretrained with the Wav2vec 2.0 loss on both the source and target domain data for domain adaptation, while the contextual encoder is frozen. Compared to a source domain finetuned model with the feature encoder being frozen during training, we find that replacing the frozen feature encoder with the adapted one provides significant WER improvements to the target domain while preserving the performance of the source domain. The effectiveness of SOA is examined on various low resource or domain mismatched ASR settings, including adult-child and clean-noisy speech.more » « less
-
Audio-based human activity recognition (HAR) is very popular because many human activities have unique sound signatures that can be detected using machine learning (ML) approaches. These audio-based ML HAR pipelines often use common featurization techniques, such as extracting various statistical and spectral features by converting time domain signals to the frequency domain (using an FFT) and using them to train ML models. Some of these approaches also claim privacy benefits by preventing the identification of human speech. However, recent deep learning-based automatic speech recognition (ASR) models pose new privacy challenges to these featurization techniques. In this paper, we systematically evaluate various featurization approaches for audio data, assessing their privacy risks through metrics like speech intelligibility (PER and WER) while considering the utility tradeoff in terms of ML-based activity recognition accuracy. Our findings reveal the susceptibility of these approaches to speech content recovery when exposed to recent ASR models, especially under re-tuning or retraining conditions. Notably, fine-tuned ASR models achieved an average Phoneme Error Rate (PER) of 39.99% and Word Error Rate (WER) of 44.43% in speech recognition for these approaches. To overcome these privacy concerns, we propose Kirigami, a lightweight machine learning-based audio speech filter that removes human speech segments reducing the efficacy of ASR models (70.48% PER and 101.40% WER) while also maintaining HAR accuracy (76.0% accuracy). We show that Kirigami can be implemented on common edge microcontrollers with limited computational capabilities and memory, providing a path to deployment on a variety of IoT devices. Finally, we conducted a real-world user study and showed the robustness of Kirigami on a laptop and an ARM Cortex-M4F microcontroller under three different background noises.more » « less
-
Modeling cross-lingual speech emotion recognition (SER) has become more prevalent because of its diverse applications. Existing studies have mostly focused on technical approaches that adapt the feature, domain, or label across languages, without considering in detail the similarities be- tween the languages. This study focuses on domain adaptation in cross-lingual scenarios using phonetic constraints. This work is framed in a twofold manner. First, we analyze emotion-specific phonetic commonality across languages by identifying common vowels that are useful for SER modeling. Second, we leverage these common vowels as an anchoring mechanism to facilitate cross-lingual SER. We consider American English and Taiwanese Mandarin as a case study to demonstrate the potential of our approach. This work uses two in-the-wild natural emotional speech corpora: MSP-Podcast (American English), and BIIC-Podcast (Taiwanese Mandarin). The proposed unsupervised cross-lingual SER model using these phonetical anchors outperforms the baselines with a 58.64% of unweighted average recall (UAR).more » « less
-
Speech recognition by both humans and machines frequently fails in non-optimal yet common situations. For example, word recognition error rates for second-language (L2) speech can be high, especially under conditions involving background noise. At the same time, both human and machine speech recognition sometimes shows remarkable robustness against signal- and noise-related degradation. Which acoustic features of speech explain this substantial variation in intelligibility? Current approaches align speech to text to extract a small set of pre-defined spectro-temporal properties from specific sounds in particular words. However, variation in these properties leaves much cross-talker variation in intelligibility unexplained. We examine an alternative approach utilizing a perceptual similarity space acquired using self-supervised learning. This approach encodes distinctions between speech samples without requiring pre-defined acoustic features or speech-to-text alignment. We show that L2 English speech samples are less tightly clustered in the space than L1 samples reflecting variability in English proficiency among L2 talkers. Critically, distances in this similarity space are perceptually meaningful: L1 English listeners have lower recognition accuracy for L2 speakers whose speech is more distant in the space from L1 speech. These results indicate that perceptual similarity may form the basis for an entirely new speech and language analysis approach.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1109/TASLP.2022.3233238
