S. Umesh

Sreyan Ghosh, Ashish Seth, and Deepak Mittal et al.

Inspired by the recent progress in self-supervised learning for computer vision, in this paper we introduce DeLoRes, a new general-purpose audio representation learning approach. Our main objective is to make our network learn representations in a resource-constrained setting (both data and compute), that can generalize well across a diverse set of downstream tasks. Inspired from the Barlow Twins objective function, we propose to learn embeddings that are invariant to distortions of an input audio sample, while making sure that they contain non-redundant information about the sample. To achieve this, we measure the cross-correlation matrix between the outputs of two identical networks fed with distorted versions of an audio segment sampled from an audio file and make it as close to the identity matrix as possible. We use a combination of a small subset of the large-scale AudioSet dataset and FSD50K for self-supervised learning and are able to learn with less than half the parameters compared to state-of-the-art algorithms. For evaluation, we transfer these learned representations to 9 downstream classification tasks, including speech, music, and animal sounds, and show competitive results under different evaluation setups. In addition to being simple and intuitive, our pre-training algorithm is amenable to compute through its inherent nature of construction and does not require careful implementation details to avoid trivial or degenerate solutions. Furthermore, we conduct ablation studies on our results and make all our code and pre-trained models publicly available https://github.com/Speech-Lab-IITM/DeLoRes.

Vasista Sai Lodagala, Sreyan Ghosh, S. Umesh

In this paper, we propose a new Self-Supervised Learning (SSL) algorithm called data2vec-aqc, for speech representation learning from unlabeled speech data. Our goal is to improve SSL for speech in domains where both unlabeled and labeled data are limited. Building on the recently introduced data2vec, we introduce additional modules to the data2vec framework that leverage the benefit of data augmentations, quantized representations, and clustering. The interaction between these modules helps solve the cross-contrastive loss as an additional self-supervised objective. data2vec-aqc achieves up to 14.1% and 20.9% relative WER improvement over the existing state-of-the-art data2vec system over the test-clean and test-other sets, respectively of LibriSpeech, without the use of any language model (LM). Our proposed model also achieves up to 17.8\% relative WER gains over the baseline data2vec when fine-tuned on a subset of the Switchboard dataset. Code: https://github.com/Speech-Lab-IITM/data2vec-aqc.

Sreyan Ghosh, Ashish Seth, S. Umesh et al.

We present Multiscale Audio Spectrogram Transformer (MAST) for audio classification, which brings the concept of multiscale feature hierarchies to the Audio Spectrogram Transformer (AST). Given an input audio spectrogram, we first patchify and project it into an initial temporal resolution and embedding dimension, post which the multiple stages in MAST progressively expand the embedding dimension while reducing the temporal resolution of the input. We use a pyramid structure that allows early layers of MAST operating at a high temporal resolution but low embedding space to model simple low-level acoustic information and deeper temporally coarse layers to model high-level acoustic information with high-dimensional embeddings. We also extend our approach to present a new Self-Supervised Learning (SSL) method called SS-MAST, which calculates a symmetric contrastive loss between latent representations from a student and a teacher encoder, leveraging patch-drop, a novel audio augmentation approach that we introduce. In practice, MAST significantly outperforms AST by an average accuracy of 3.4% across 8 speech and non-speech tasks from the LAPE Benchmark, achieving state-of-the-art results on keyword spotting in Speech Commands. Additionally, our proposed SS-MAST achieves an absolute average improvement of 2.6% over the previously proposed SSAST.

Ashish Seth, Sreyan Ghosh, S. Umesh et al.

We present a new Self-Supervised Learning (SSL) approach to pre-train encoders on unlabeled audio data that reduces the need for large amounts of labeled data for audio and speech classification. Our primary aim is to learn audio representations that can generalize across a large variety of speech and non-speech tasks in a low-resource un-labeled audio pre-training setting. Inspired by the recent success of clustering and contrasting learning paradigms for SSL-based speech representation learning, we propose SLICER (Symmetrical Learning of Instance and Cluster-level Efficient Representations), which brings together the best of both clustering and contrasting learning paradigms. We use a symmetric loss between latent representations from student and teacher encoders and simultaneously solve instance and cluster-level contrastive learning tasks. We obtain cluster representations online by just projecting the input spectrogram into an output subspace with dimensions equal to the number of clusters. In addition, we propose a novel mel-spectrogram augmentation procedure, k-mix, based on mixup, which does not require labels and aids unsupervised representation learning for audio. Overall, SLICER achieves state-of-the-art results on the LAPE Benchmark \cite{9868132}, significantly outperforming DeLoRes-M and other prior approaches, which are pre-trained on $10\times$ larger of unsupervised data. We will make all our codes available on GitHub.

Vasista Sai Lodagala, Sreyan Ghosh, S. Umesh

While Self-Supervised Learning has helped reap the benefit of the scale from the available unlabeled data, the learning paradigms are continuously being bettered. We present a new pre-training strategy named ccc-wav2vec 2.0, which uses clustering and an augmentation-based cross-contrastive loss as its self-supervised objective. Through the clustering module, we scale down the influence of those negative examples that are highly similar to the positive. The Cross-Contrastive loss is computed between the encoder output of the original sample and the quantizer output of its augmentation and vice-versa, bringing robustness to the pre-training strategy. ccc-wav2vec 2.0 achieves up to 15.6% and 12.7% relative WER improvement over the baseline wav2vec 2.0 on the test-clean and test-other sets, respectively, of LibriSpeech, without the use of any language model. The proposed method also achieves up to 14.9% relative WER improvement over the baseline wav2vec 2.0 when fine-tuned on Switchboard data. We make all our codes publicly available on GitHub.

A Arunkumar, Vrunda N Sukhadia, S. Umesh

Self-supervised learning (SSL) based models have been shown to generate powerful representations that can be used to improve the performance of downstream speech tasks. Several state-of-the-art SSL models are available, and each of these models optimizes a different loss which gives rise to the possibility of their features being complementary. This paper proposes using an ensemble of such SSL representations and models, which exploits the complementary nature of the features extracted by the various pretrained models. We hypothesize that this results in a richer feature representation and shows results for the ASR downstream task. To this end, we use three SSL models that have shown excellent results on ASR tasks, namely HuBERT, Wav2vec2.0, and WaveLM. We explore the ensemble of models fine-tuned for the ASR task and the ensemble of features using the embeddings obtained from the pre-trained models for a downstream ASR task. We get improved performance over individual models and pre-trained features using Librispeech(100h) and WSJ dataset for the downstream tasks.

Ashish Seth, Sreyan Ghosh, S. Umesh et al.

In this paper, we introduce UnFuSeD, a novel approach to leverage self-supervised learning and reduce the need for large amounts of labeled data for audio classification. Unlike prior works, which directly fine-tune a self-supervised pre-trained encoder on a target dataset, we use the encoder to generate pseudo-labels for unsupervised fine-tuning before the actual fine-tuning step. We first train an encoder using a novel self-supervised learning algorithm (SSL) on an unlabeled audio dataset. Then, we use that encoder to generate pseudo-labels on our target task dataset via clustering the extracted representations. These pseudo-labels are then used to guide self-distillation on a randomly initialized model, which we call unsupervised fine-tuning. Finally, the resultant encoder is then fine-tuned on our target task dataset. Through UnFuSeD, we propose the first system that moves away from generic SSL paradigms in literature, which pre-train and fine-tune the same encoder, and present a novel self-distillation-based system to leverage SSL pre-training for low-resource audio classification. In practice, UnFuSeD achieves state-of-the-art results on the LAPE Benchmark, significantly outperforming all our baselines. Additionally, UnFuSeD allows us to achieve this at a 40% reduction in the number of parameters over the previous state-of-the-art system. We make all our codes publicly available.

Sreyan Ghosh, Sonal Kumar, Yaman Kumar Singla et al.

Existing approaches in disfluency detection focus on solving a token-level classification task for identifying and removing disfluencies in text. Moreover, most works focus on leveraging only contextual information captured by the linear sequences in text, thus ignoring the structured information in text which is efficiently captured by dependency trees. In this paper, building on the span classification paradigm of entity recognition, we propose a novel architecture for detecting disfluencies in transcripts from spoken utterances, incorporating both contextual information through transformers and long-distance structured information captured by dependency trees, through graph convolutional networks (GCNs). Experimental results show that our proposed model achieves state-of-the-art results on the widely used English Switchboard for disfluency detection and outperforms prior-art by a significant margin. We make all our codes publicly available on GitHub (https://github.com/Sreyan88/Disfluency-Detection-with-Span-Classification)

Vishwas M. Shetty, Metilda Sagaya Mary N J, S. Umesh

End-to-end models are fast replacing the conventional hybrid models in automatic speech recognition. Transformer, a sequence-to-sequence model, based on self-attention popularly used in machine translation tasks, has given promising results when used for automatic speech recognition. This paper explores different ways of incorporating speaker information at the encoder input while training a transformer-based model to improve its speech recognition performance. We present speaker information in the form of speaker embeddings for each of the speakers. We experiment using two types of speaker embeddings: x-vectors and novel s-vectors proposed in our previous work. We report results on two datasets a) NPTEL lecture database and b) Librispeech 500-hour split. NPTEL is an open-source e-learning portal providing lectures from top Indian universities. We obtain improvements in the word error rate over the baseline through our approach of integrating speaker embeddings into the model.

Ashish Seth, Sreyan Ghosh, S. Umesh et al.

Continued self-supervised (SSL) pre-training for adapting existing SSL models to the target domain has shown to be extremely effective for low-resource Automatic Speech Recognition (ASR). This paper proposes Stable Distillation, a simple and novel approach for SSL-based continued pre-training that boosts ASR performance in the target domain where both labeled and unlabeled data are limited. Stable Distillation employs self-distillation as regularization for continued pre-training, alleviating the over-fitting issue, a common problem continued pre-training faces when the source and target domains differ. Specifically, first, we perform vanilla continued pre-training on an initial SSL pre-trained model on the target domain ASR dataset and call it the teacher. Next, we take the same initial pre-trained model as a student to perform continued pre-training while enforcing its hidden representations to be close to that of the teacher (via MSE loss). This student is then used for downstream ASR fine-tuning on the target dataset. In practice, Stable Distillation outperforms all our baselines by 0.8 - 7 WER when evaluated in various experimental settings.

Ashish Seth, Sreyan Ghosh, S. Umesh et al.

Continued pre-training (CP) offers multiple advantages, like target domain adaptation and the potential to exploit the continuous stream of unlabeled data available online. However, continued pre-training on out-of-domain distributions often leads to catastrophic forgetting of previously acquired knowledge, leading to sub-optimal ASR performance. This paper presents FusDom, a simple and novel methodology for SSL-based continued pre-training. FusDom learns speech representations that are robust and adaptive yet not forgetful of concepts seen in the past. Instead of solving the SSL pre-text task on the output representations of a single model, FusDom leverages two identical pre-trained SSL models, a teacher and a student, with a modified pre-training head to solve the CP SSL pre-text task. This head employs a cross-attention mechanism between the representations of both models while only the student receives gradient updates and the teacher does not. Finally, the student is fine-tuned for ASR. In practice, FusDom outperforms all our baselines across settings significantly, with WER improvements in the range of 0.2 WER - 7.3 WER in the target domain while retaining the performance in the earlier domain.

Arjun Gangwar, Kaousheik Jayakumar, S. Umesh

This paper describes the systems developed by SPRING Lab, Indian Institute of Technology Madras, for the ASRU MADASR 2.0 challenge. The systems developed focuses on adapting ASR systems to improve in predicting the language and dialect of the utterance among 8 languages across 33 dialects. We participated in Track 1 and Track 2, which restricts the use of additional data and develop from-the-scratch multilingual systems. We presented a novel training approach using Multi-Decoder architecture with phonemic Common Label Set (CLS) as intermediate representation. It improved the performance over the baseline (in the CLS space). We also discuss various methods used to retain the gain obtained in the phonemic space while converting them back to the corresponding grapheme representations. Our systems beat the baseline in 3 languages (Track 2) in terms of WER/CER and achieved the highest language ID and dialect ID accuracy among all participating teams (Track 2).

Advait Joglekar, Divyanshu Singh, Rooshil Rohit Bhatia et al.

Voice Conversion research in recent times has increasingly focused on improving the zero-shot capabilities of existing methods. Despite remarkable advancements, current architectures still tend to struggle in zero-shot cross-lingual settings. They are also often unable to generalize for speakers of unseen languages and accents. In this paper, we adopt a simple yet effective approach that combines discrete speech representations from self-supervised models with a non-autoregressive Diffusion-Transformer based conditional flow matching speech decoder. We show that this architecture allows us to train a voice-conversion model in a purely textless, self-supervised fashion. Our technique works without requiring multiple encoders to disentangle speech features. Our model also manages to excel in zero-shot cross-lingual settings even for unseen languages. For Demo: https://ez-vc.github.io/EZ-VC-Demo/

Kaousheik Jayakumar, Vrunda N. Sukhadia, A Arunkumar et al.

Building a multilingual Automated Speech Recognition (ASR) system in a linguistically diverse country like India can be a challenging task due to the differences in scripts and the limited availability of speech data. This problem can be solved by exploiting the fact that many of these languages are phonetically similar. These languages can be converted into a Common Label Set (CLS) by mapping similar sounds to common labels. In this paper, new approaches are explored and compared to improve the performance of CLS based multilingual ASR model. Specific language information is infused in the ASR model by giving Language ID or using CLS to Native script converter on top of the CLS Multilingual model. These methods give a significant improvement in Word Error Rate (WER) compared to the CLS baseline. These methods are further tried on out-of-distribution data to check their robustness.

Ashish Seth, Lodagala V S V Durga Prasad, Sreyan Ghosh et al.

Self-supervised learning (SSL) to learn high-level speech representations has been a popular approach to building Automatic Speech Recognition (ASR) systems in low-resource settings. However, the common assumption made in literature is that a considerable amount of unlabeled data is available for the same domain or language that can be leveraged for SSL pre-training, which we acknowledge is not feasible in a real-world setting. In this paper, as part of the Interspeech Gram Vaani ASR challenge, we try to study the effect of domain, language, dataset size, and other aspects of our upstream pre-training SSL data on the final performance low-resource downstream ASR task. We also build on the continued pre-training paradigm to study the effect of prior knowledge possessed by models trained using SSL. Extensive experiments and studies reveal that the performance of ASR systems is susceptible to the data used for SSL pre-training. Their performance improves with an increase in similarity and volume of pre-training data. We believe our work will be helpful to the speech community in building better ASR systems in low-resource settings and steer research towards improving generalization in SSL-based pre-training for speech systems.

Lodagala V S V Durga Prasad, Sreyan Ghosh, S. Umesh

While self-supervised speech representation learning (SSL) models serve a variety of downstream tasks, these models have been observed to overfit to the domain from which the unlabelled data originates. To alleviate this issue, we propose PADA (Pruning Assisted Domain Adaptation) and zero out redundant weights from models pre-trained on large amounts of out-of-domain (OOD) data. Intuitively, this helps to make space for the target-domain ASR finetuning. The redundant weights can be identified through various pruning strategies which have been discussed in detail as a part of this work. Specifically, we investigate the effect of the recently discovered Task-Agnostic and Task-Aware pruning on PADA and propose a new pruning paradigm based on the latter, which we call Cross-Domain Task-Aware Pruning (CD-TAW). CD-TAW obtains the initial pruning mask from a well fine-tuned OOD model, which makes it starkly different from the rest of the pruning strategies discussed in the paper. Our proposed CD-TAW methodology achieves up to 20.6% relative WER improvement over our baseline when fine-tuned on a 2-hour subset of Switchboard data without language model (LM) decoding. Furthermore, we conduct a detailed analysis to highlight the key design choices of our proposed method.

Vrunda N. Sukhadia, S. Umesh

In this paper, we investigate domain adaptation for low-resource Automatic Speech Recognition (ASR) of target-domain data, when a well-trained ASR model trained with a large dataset is available. We argue that in the encoder-decoder framework, the decoder of the well-trained ASR model is largely tuned towards the source-domain, hurting the performance of target-domain models in vanilla transfer-learning. On the other hand, the encoder layers of the well-trained ASR model mostly capture the acoustic characteristics. We, therefore, propose to use the embeddings tapped from these encoder layers as features for a downstream Conformer target-domain model and show that they provide significant improvements. We do ablation studies on which encoder layer is optimal to tap the embeddings, as well as the effect of freezing or updating the well-trained ASR model's encoder layers. We further show that applying Spectral Augmentation (SpecAug) on the proposed features (this is in addition to default SpecAug on input spectral features) provides a further improvement on the target-domain performance. For the LibriSpeech-100-clean data as target-domain and SPGI-5000 as a well-trained model, we get 30% relative improvement over baseline. Similarly, with WSJ data as target-domain and LibriSpeech-960 as a well-trained model, we get 50% relative improvement over baseline.

Sreyan Ghosh, Sandesh V Katta, Ashish Seth et al.

We introduce DECAR, a self-supervised pre-training approach for learning general-purpose audio representations. Our system is based on clustering: it utilizes an offline clustering step to provide target labels that act as pseudo-labels for solving a prediction task. We develop on top of recent advances in self-supervised learning for computer vision and design a lightweight, easy-to-use self-supervised pre-training scheme. We pre-train DECAR embeddings on a balanced subset of the large-scale Audioset dataset and transfer those representations to 9 downstream classification tasks, including speech, music, animal sounds, and acoustic scenes. Furthermore, we conduct ablation studies identifying key design choices and also make all our code and pre-trained models publicly available.

Sreyan Ghosh, Samden Lepcha, S Sakshi et al.

Toxic speech, also known as hate speech, is regarded as one of the crucial issues plaguing online social media today. Most recent work on toxic speech detection is constrained to the modality of text and written conversations with very limited work on toxicity detection from spoken utterances or using the modality of speech. In this paper, we introduce a new dataset DeToxy, the first publicly available toxicity annotated dataset for the English language. DeToxy is sourced from various openly available speech databases and consists of over 2 million utterances. We believe that our dataset would act as a benchmark for the relatively new and un-explored Spoken Language Processing task of detecting toxicity from spoken utterances and boost further research in this space. Finally, we also provide strong unimodal baselines for our dataset and compare traditional two-step and E2E approaches. Our experiments show that in the case of spoken utterances, text-based approaches are largely dependent on gold human-annotated transcripts for their performance and also suffer from the problem of keyword bias. However, the presence of speech files in DeToxy helps facilitates the development of E2E speech models which alleviate both the above-stated problems by better capturing speech clues.

D. S. Pavan Kumar, N. Vishnu Prasad, Vikas Joshi et al.

In this paper, a modification to the training process of the popular SPLICE algorithm has been proposed for noise robust speech recognition. The modification is based on feature correlations, and enables this stereo-based algorithm to improve the performance in all noise conditions, especially in unseen cases. Further, the modified framework is extended to work for non-stereo datasets where clean and noisy training utterances, but not stereo counterparts, are required. Finally, an MLLR-based computationally efficient run-time noise adaptation method in SPLICE framework has been proposed. The modified SPLICE shows 8.6% absolute improvement over SPLICE in Test C of Aurora-2 database, and 2.93% overall. Non-stereo method shows 10.37% and 6.93% absolute improvements over Aurora-2 and Aurora-4 baseline models respectively. Run-time adaptation shows 9.89% absolute improvement in modified framework as compared to SPLICE for Test C, and 4.96% overall w.r.t. standard MLLR adaptation on HMMs.