Shankar M. Venkatesan

Sergio Burdisso, Esaú Villatoro-Tello, Andrés Carofilis et al.

Adapting automatic speech recognition (ASR) systems based on large language models (LLMs) to new domains using text-only data is a significant yet underexplored challenge. Standard fine-tuning of the LLM on target-domain text often disrupts the critical alignment between speech and text modalities learned by the projector, degrading performance. We introduce a novel text-only adaptation method that emulates the audio projection task by treating it as a text denoising task. Our approach thus trains the LLM to recover clean transcripts from noisy inputs. This process effectively adapts the model to a target domain while preserving cross-modal alignment. Our solution is lightweight, requiring no architectural changes or additional parameters. Extensive evaluation on two datasets demonstrates up to 22.1% relative improvement, outperforming recent state-of-the-art text-only adaptation methods.

Pradeep Rangappa, Andres Carofilis, Jeena Prakash et al.

Fine-tuning pretrained ASR models for specific domains is challenging for small organizations with limited labeled data and computational resources. Here, we explore different data selection pipelines and propose a robust approach that improves ASR adaptation by filtering pseudo-labels generated using Whisper (encoder-decoder) and Zipformer (transducer) models. Our approach integrates multiple selection strategies -- including word error rate (WER) prediction, named entity recognition (NER), and character error rate (CER) analysis -- to extract high-quality training segments. We evaluate our method on Whisper and Zipformer using a 7500-hour baseline, comparing it to a CER-based approach relying on hypotheses from three ASR systems. Fine-tuning on 7500 hours of pseudo-labeled call center data achieves 12.3% WER, while our filtering reduces the dataset to 100 hours (1.4%) with similar performance; a similar trend is observed on Fisher English.

Jeena Prakash, Blessingh Kumar, Kadri Hacioglu et al.

Automatic speech recognition (ASR) models rely on high-quality transcribed data for effective training. Generating pseudo-labels for large unlabeled audio datasets often relies on complex pipelines that combine multiple ASR outputs through multi-stage processing, leading to error propagation, information loss and disjoint optimization. We propose a unified multi-ASR prompt-driven framework using postprocessing by either textual or speech-based large language models (LLMs), replacing voting or other arbitration logic for reconciling the ensemble outputs. We perform a comparative study of multiple architectures with and without LLMs, showing significant improvements in transcription accuracy compared to traditional methods. Furthermore, we use the pseudo-labels generated by the various approaches to train semi-supervised ASR models for different datasets, again showing improved performance with textual and speechLLM transcriptions compared to baselines.

Bidisha Sharma, Karthik Pandia Durai, Shankar Venkatesan et al.

There has been increasing interest in unifying streaming and non-streaming automatic speech recognition (ASR) models to reduce development, training, and deployment costs. We present a unified framework that trains a single end-to-end ASR model for both streaming and non-streaming applications, leveraging future context information. We propose to use dynamic right-context through the chunked attention masking in the training of zipformer-based ASR models. We demonstrate that using right-context is more effective in zipformer models compared to other conformer models due to its multi-scale nature. We analyze the effect of varying the number of right-context frames on accuracy and latency of the streaming ASR models. We use Librispeech and large in-house conversational datasets to train different versions of streaming and non-streaming models and evaluate them in a production grade server-client setup across diverse testsets of different domains. The proposed strategy reduces word error by relative 7.9\% with a small degradation in user-perceived latency. By adding more right-context frames, we are able to achieve streaming performance close to that of non-streaming models. Our approach also allows flexible control of the latency-accuracy tradeoff according to customers requirements.

Andres Carofilis, Pradeep Rangappa, Srikanth Madikeri et al.

Fine-tuning pretrained ASR models for specific domains is challenging when labeled data is scarce. But unlabeled audio and labeled data from related domains are often available. We propose an incremental semi-supervised learning pipeline that first integrates a small in-domain labeled set and an auxiliary dataset from a closely related domain, achieving a relative improvement of 4% over no auxiliary data. Filtering based on multi-model consensus or named entity recognition (NER) is then applied to select and iteratively refine pseudo-labels, showing slower performance saturation compared to random selection. Evaluated on the multi-domain Wow call center and Fisher English corpora, it outperforms single-step fine-tuning. Consensus-based filtering outperforms other methods, providing up to 22.3% relative improvement on Wow and 24.8% on Fisher over single-step fine-tuning with random selection. NER is the second-best filter, providing competitive performance at a lower computational cost.

Paritosh Mittal, Shankar M Venkatesan, Viswanath Veera et al.

People capture memorable images of events and exhibits that are often occluded by a wire mesh loosely termed as fence. Recent works in removing fence have limited performance due to the difficulty in initial fence segmentation. This work aims to accurately segment fence using a novel fence guidance mask (FM) generated from stereo image pair. This binary guidance mask contains deterministic cues about the structure of fence and is given as additional input to the deep fence estimation model. We also introduce a directional connectivity loss (DCL), which is used alongside adversarial loss to precisely detect thin wires. Experimental results obtained on real world scenarios demonstrate the superiority of proposed method over state-of-the-art techniques.

Rohith Aralikatti, Sharad Roy, Abhinav Thanda et al.

Under noisy conditions, speech recognition systems suffer from high Word Error Rates (WER). In such cases, information from the visual modality comprising the speaker lip movements can help improve the performance. In this work, we propose novel methods to fuse information from audio and visual modalities at inference time. This enables us to train the acoustic and visual models independently. First, we train separate RNN-HMM based acoustic and visual models. A common WFST generated by taking a special union of the HMM components is used for decoding using a modified Viterbi algorithm. Second, we train separate seq2seq acoustic and visual models. The decoding step is performed simultaneously for both modalities using shallow fusion while maintaining a common hypothesis beam. We also present results for a novel seq2seq fusion without the weighing parameter. We present results at varying SNR and show that our methods give significant improvements over acoustic-only WER.

Dilip Kumar Margam, Rohith Aralikatti, Tanay Sharma et al.

In recent years, deep learning based machine lipreading has gained prominence. To this end, several architectures such as LipNet, LCANet and others have been proposed which perform extremely well compared to traditional lipreading DNN-HMM hybrid systems trained on DCT features. In this work, we propose a simpler architecture of 3D-2D-CNN-BLSTM network with a bottleneck layer. We also present analysis of two different approaches for lipreading on this architecture. In the first approach, 3D-2D-CNN-BLSTM network is trained with CTC loss on characters (ch-CTC). Then BLSTM-HMM model is trained on bottleneck lip features (extracted from 3D-2D-CNN-BLSTM ch-CTC network) in a traditional ASR training pipeline. In the second approach, same 3D-2D-CNN-BLSTM network is trained with CTC loss on word labels (w-CTC). The first approach shows that bottleneck features perform better compared to DCT features. Using the second approach on Grid corpus' seen speaker test set, we report $1.3\%$ WER - a $55\%$ improvement relative to LCANet. On unseen speaker test set we report $8.6\%$ WER which is $24.5\%$ improvement relative to LipNet. We also verify the method on a second dataset of $81$ speakers which we collected. Finally, we also discuss the effect of feature duplication on BLSTM-HMM model performance.

Rohith Aralikatti, Dilip Margam, Tanay Sharma et al.

This paper demonstrates two novel methods to estimate the global SNR of speech signals. In both methods, Deep Neural Network-Hidden Markov Model (DNN-HMM) acoustic model used in speech recognition systems is leveraged for the additional task of SNR estimation. In the first method, the entropy of the DNN-HMM output is computed. Recent work on bayesian deep learning has shown that a DNN-HMM trained with dropout can be used to estimate model uncertainty by approximating it as a deep Gaussian process. In the second method, this approximation is used to obtain model uncertainty estimates. Noise specific regressors are used to predict the SNR from the entropy and model uncertainty. The DNN-HMM is trained on GRID corpus and tested on different noise profiles from the DEMAND noise database at SNR levels ranging from -10 dB to 30 dB.

Abhinav Thanda, Shankar M Venkatesan

Multi-task learning (MTL) involves the simultaneous training of two or more related tasks over shared representations. In this work, we apply MTL to audio-visual automatic speech recognition(AV-ASR). Our primary task is to learn a mapping between audio-visual fused features and frame labels obtained from acoustic GMM/HMM model. This is combined with an auxiliary task which maps visual features to frame labels obtained from a separate visual GMM/HMM model. The MTL model is tested at various levels of babble noise and the results are compared with a base-line hybrid DNN-HMM AV-ASR model. Our results indicate that MTL is especially useful at higher level of noise. Compared to base-line, upto 7\% relative improvement in WER is reported at -3 SNR dB

Abhinav Thanda, Shankar M Venkatesan

In this work, we propose a training algorithm for an audio-visual automatic speech recognition (AV-ASR) system using deep recurrent neural network (RNN).First, we train a deep RNN acoustic model with a Connectionist Temporal Classification (CTC) objective function. The frame labels obtained from the acoustic model are then used to perform a non-linear dimensionality reduction of the visual features using a deep bottleneck network. Audio and visual features are fused and used to train a fusion RNN. The use of bottleneck features for visual modality helps the model to converge properly during training. Our system is evaluated on GRID corpus. Our results show that presence of visual modality gives significant improvement in character error rate (CER) at various levels of noise even when the model is trained without noisy data. We also provide a comparison of two fusion methods: feature fusion and decision fusion.