Research work on the design of robust multimodal speech recognition systems making use of acoustic and visual cues, extracted using the relatively noise robust alternate speech sensors is gaining interest in recent times among the speech processing research fraternity. The primary objective of this work is to study the exclusive influence of Lombard effect on the automatic recognition of the confusable syllabic consonant-vowel units of Hindi language, as a step towards building robust multimodal ASR systems in adverse environments in the context of Indian languages which are syllabic in nature. The dataset for this work comprises the confusable 145 consonant-vowel (CV) syllabic units of Hindi language recorded simultaneously using three modalities that capture the acoustic and visual speech cues, namely normal acoustic microphone (NM), throat microphone (TM) and a camera that captures the associated lip movements. The Lombard effect is induced by feeding crowd noise into the speaker’s headphone while recording. Convolutional Neural Network (CNN) models are built to categorise the CV units based on their place of articulation (POA), manner of articulation (MOA), and vowels (under clean and Lombard conditions). For validation purpose, corresponding Hidden Markov Models (HMM) are also built and tested. Unimodal Automatic Speech Recognition (ASR) systems built using each of the three speech cues from Lombard speech show a loss in recognition of MOA and vowels while POA gets a boost in all the systems due to Lombard effect. Combining the three complimentary speech cues to build bimodal and trimodal ASR systems shows that the recognition loss due to Lombard effect for MOA and vowels reduces compared to the unimodal systems, while the POA recognition is still better due to Lombard effect. A bimodal system is proposed using only alternate acoustic and visual cues which gives a better discrimination of the place and manner of articulation than even standard ASR system. Among the multimodal ASR systems studied, the proposed trimodal system based on Lombard speech gives the best recognition accuracy of 98%, 95%, and 76% for the vowels, MOA and POA, respectively, with an average improvement of 36% over the unimodal ASR systems and 9% improvement over the bimodal ASR systems.

Speech recognition system extract the textual data from the speech signal. The research in speech recognition domain is challenging due to the large variabilities involved with the speech signal. Variety of signal processing and machine learning techniques have been explored to achieve better recognition accuracy. Speech is highly non-stationary in nature and therefore analysis is carried out by considering short time-domain window or frame. In the speech recognition task, cepstral (Mel frequency cepstral coefficients (MFCC)) features are commonly used and are extracted for short time-frame. The effectiveness of features depend upon duration of the time-window chosen. The present study is aimed at investigation of optimal time-window duration for extraction of cepstral features in the context of speech recognition task. A speaker independent speech recognition system for the Kannada language has been considered for the analysis. In the current work, speech utterances of Kannada news corpus recorded from different speakers have been used to create speech database. The hidden Markov tool kit (HTK) has been used to implement the speech recognition system. The MFCC along with their first and second derivative coefficients are considered as feature vectors. Pronunciation dictionary required for the study has been built manually for mono-phone system. Experiments have been carried out and results have been analyzed for different time-window lengths. The overlapping Hamming window has been considered in this study. The best average word recognition accuracy of 61.58% has been obtained for a window length of 110 msec duration. This recognition accuracy is comparable with the similar work found in literature. The experiments have shown that best word recognition performance can be achieved by tuning the window length to its optimum value.