Research in Automatic Speech Recognition (ASR) has witnessed a steep improvement in the past decade (especially for English language) where the variety and amount of training data available is huge. In this work, we develop an ASR and Keyword Search (KWS) system for Manipuri, a low-resource Indian Language. Manipuri (also known as Meitei), is a Tibeto-Burman language spoken predominantly in Manipur (a northeastern state of India). We collect and transcribe telephonic read speech data of 90+ hours from 300+ speakers for the ASR task. Both state-of-the-art Gaussian Mixture-Hidden Markov Model (GMM-HMM) and Deep Neural Network-Hidden Markov Model (DNN-HMM) based architectures are developed as a baseline. Using the collected data, we achieve better performance using DNN-HMM systems, i.e., 13.57% WER for ASR and 7.64% EER for KWS. The KALDI speech recognition tool-kit is used for developing the systems. The Manipuri ASR system along with KWS is integrated as a visual interface for demonstration purpose. Future systems will be improved with more amount of training data and advanced forms of acoustic models and language models.
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.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.