Don't speak too fast: The impact of data bias on self-supervised speech models

Self-supervised Speech Models (S3Ms) have been proven successful in many speech downstream tasks, like ASR. However, how pre-training data affects S3Ms' downstream behavior remains an unexplored issue. In this paper, we study how pre-training data affects S3Ms by pre-training models on biased datasets targeting different factors of speech, including gender, content, and prosody, and evaluate these pre-trained S3Ms on selected downstream tasks in SUPERB Benchmark. Our experiments show that S3Ms have tolerance toward gender bias. Moreover, we find that the content of speech has little impact on the performance of S3Ms across downstream tasks, but S3Ms do show a preference toward a slower speech rate.

Toward Degradation-Robust Voice Conversion

Any-to-any voice conversion technologies convert the vocal timbre of an utterance to any speaker even unseen during training. Although there have been several state-of-the-art any-to-any voice conversion models, they were all based on clean utterances to convert successfully. However, in real-world scenarios, it is difficult to collect clean utterances of a speaker, and they are usually degraded by noises or reverberations. It thus becomes highly desired to understand how these degradations affect voice conversion and build a degradation-robust model. We report in this paper the first comprehensive study on the degradation robustness of any-to-any voice conversion. We show that the performance of state-of-the-art models nowadays was severely hampered given degraded utterances. To this end, we then propose speech enhancement concatenation and denoising training to improve the robustness. In addition to common degradations, we also consider adversarial noises, which alter the model output significantly yet are human-imperceptible. It was shown that both concatenations with off-the-shelf speech enhancement models and denoising training on voice conversion models could improve the robustness, while each of them had pros and cons.

An Exploration of Self-Supervised Pretrained Representations for End-to-End Speech Recognition

Self-supervised pretraining on speech data has achieved a lot of progress. High-fidelity representation of the speech signal is learned from a lot of untranscribed data and shows promising performance. Recently, there are several works focusing on evaluating the quality of self-supervised pretrained representations on various tasks without domain restriction, e.g. SUPERB. However, such evaluations do not provide a comprehensive comparison among many ASR benchmark corpora. In this paper, we focus on the general applications of pretrained speech representations, on advanced end-to-end automatic speech recognition (E2E-ASR) models. We select several pretrained speech representations and present the experimental results on various open-source and publicly available corpora for E2E-ASR. Without any modification of the back-end model architectures or training strategy, some of the experiments with pretrained representations, e.g., WSJ, WSJ0-2mix with HuBERT, reach or outperform current state-of-the-art (SOTA) recognition performance. Moreover, we further explore more scenarios for whether the pretraining representations are effective, such as the cross-language or overlapped speech. The scripts, configuratons and the trained models have been released in ESPnet to let the community reproduce our experiments and improve them.

Mandarin-English Code-switching Speech Recognition with Self-supervised Speech Representation Models

Code-switching (CS) is common in daily conversations where more than one language is used within a sentence. The difficulties of CS speech recognition lie in alternating languages and the lack of transcribed data. Therefore, this paper uses the recently successful self-supervised learning (SSL) methods to leverage many unlabeled speech data without CS. We show that hidden representations of SSL models offer frame-level language identity even if the models are trained with English speech only. Jointly training CTC and language identification modules with self-supervised speech representations improves CS speech recognition performance. Furthermore, using multilingual speech data for pre-training obtains the best CS speech recognition.

DistilHuBERT: Speech Representation Learning by Layer-wise Distillation of Hidden-unit BERT

Self-supervised speech representation learning methods like wav2vec 2.0 and Hidden-unit BERT (HuBERT) leverage unlabeled speech data for pre-training and offer good representations for numerous speech processing tasks. Despite the success of these methods, they require large memory and high pre-training costs, making them inaccessible for researchers in academia and small companies. Therefore, this paper introduces DistilHuBERT, a novel multi-task learning framework to distill hidden representations from a HuBERT model directly. This method reduces HuBERT's size by 75% and 73% faster while retaining most performance in ten different tasks. Moreover, DistilHuBERT required little training time and data, opening the possibilities of pre-training personal and on-device SSL models for speech.

On the Transferability of Pre-trained Language Models: A Study from Artificial Datasets

Pre-training language models (LMs) on large-scale unlabeled text data makes the model much easier to achieve exceptional downstream performance than their counterparts directly trained on the downstream tasks. In this work, we study what specific traits in the pre-training data, other than the semantics, make a pre-trained LM superior to their counterparts trained from scratch on downstream tasks. We propose to use artificially constructed datasets as the pre-training data to exclude the effect of semantics, and further control what characteristics the pre-training corpora have. By fine-tuning the pre-trained models on GLUE benchmark, we can learn how beneficial it is to transfer the knowledge from the model trained on the dataset possessing that specific trait. We define and discuss three different characteristics in the artificial dataset: 1) matching the token's uni-gram or bi-gram distribution between pre-training and downstream fine-tuning, 2) the presence of the explicit dependencies among the tokens in a sequence, 3) the length of the implicit dependencies among the tokens in a sequence. Our experiments show that the explicit dependencies in the sequences of the pre-training data are critical to the downstream performance. Our results also reveal that models achieve better downstream performance when pre-trained on a dataset with a longer range of implicit dependencies. Based on our analysis, we find that models pre-trained with artificial datasets are prone to learn spurious correlation in downstream tasks. Our work reveals that even if the LMs are not pre-trained on natural language, they still gain transferability on certain human language downstream tasks once the LMs learn to model the token dependencies in the sequences. This result helps us understand the exceptional transferability of pre-trained LMs.

Parallelized Reverse Curriculum Generation

For reinforcement learning (RL), it is challenging for an agent to master a task that requires a specific series of actions due to sparse rewards. To solve this problem, reverse curriculum generation (RCG) provides a reverse expansion approach that automatically generates a curriculum for the agent to learn. More specifically, RCG adapts the initial state distribution from the neighborhood of a goal to a distance as training proceeds. However, the initial state distribution generated for each iteration might be biased, thus making the policy overfit or slowing down the reverse expansion rate. While training RCG for actor-critic (AC) based RL algorithms, this poor generalization and slow convergence might be induced by the tight coupling between an AC pair. Therefore, we propose a parallelized approach that simultaneously trains multiple AC pairs and periodically exchanges their critics. We empirically demonstrate that this proposed approach can improve RCG in performance and convergence, and it can also be applied to other AC based RL algorithms with adapted initial state distribution.

Spotting adversarial samples for speaker verification by neural vocoders

Automatic speaker verification (ASV), one of the most important technology for biometric identification, has been widely adopted in security-critical applications, including transaction authentication and access control. However, previous work has shown that ASV is seriously vulnerable to recently emerged adversarial attacks, yet effective countermeasures against them are limited. In this paper, we adopt neural vocoders to spot adversarial samples for ASV. We use the neural vocoder to re-synthesize audio and find that the difference between the ASV scores for the original and re-synthesized audio is a good indicator for discrimination between genuine and adversarial samples. This effort is, to the best of our knowledge, among the first to pursue such a technical direction for detecting adversarial samples for ASV, and hence there is a lack of established baselines for comparison. Consequently, we implement the Griffin-Lim algorithm as the detection baseline. The proposed approach achieves effective detection performance that outperforms all the baselines in all the settings. We also show that the neural vocoder adopted in the detection framework is dataset-independent. Our codes will be made open-source for future works to do comparison.

Multi-accent Speech Separation with One Shot Learning

Speech separation is a problem in the field of speech processing that has been studied in full swing recently. However, there has not been much work studying a multi-accent speech separation scenario. Unseen speakers with new accents and noise aroused the domain mismatch problem which cannot be easily solved by conventional joint training methods. Thus, we applied MAML and FOMAML to tackle this problem and obtained higher average Si-SNRi values than joint training on almost all the unseen accents. This proved that these two methods do have the ability to generate well-trained parameters for adapting to speech mixtures of new speakers and accents. Furthermore, we found out that FOMAML obtains similar performance compared to MAML while saving a lot of time.