Microphone arrays and neural networks for robust speech recognition

Abstract

This paper explores use of synergistically-integrated systems of microphone arrays and neural networks for robust speech recognition in variable acoustic environments, where the user must not be encumbered by microphone equipment. Existing speech recognizers work best for "high-quality close-talking speech." Performance of these recognizers is typically degraded by environmental interference and mismatch in training conditions and testing conditions. It is found that use of microphone arrays and neural network processors can elevate the recognition performance of existing speech recognizers in an adverse acoustic environment, thus avoiding the need to retrain the recognizer, a complex and tedious task. We also present results showing that a system of microphone arrays and neural networks can achieve a higher word recognition accuracy in an unmatched training/testing condition than that obtained with a retrained speech recognizer using array speech for both training and testing, i.e., a matched training/testing condition.

A survey of hybrid ANN/HMM models for automatic speech recognition

“…Related problems, which have to deal with system robustness, concern the electronic transducer, namely the microphone(s) used to capture the voice signal: the microphone used for testing may di!er from that used to collect the training data, or it may no longer be close-mouth but placed at a certain distance from the talker(s). A microphone array [35,19] is sometimes needed to allow hands-free recognition of people moving within a room, and so on. Model adaptation techniques usually aim at building an acoustic front-end to a pre-trained HMM-based recognizer, or to tune the parameters of the latter, in order to increase recognition performance whenever acoustic conditions are changed with respect to those which held during the training, using a limited amount of acoustic material collected in the new conditions and avoiding a full re-training of the whole system.…”

A practical methodology for speech source localization with microphone arrays

“…Existing array systems have been used in a number of applications. These include teleconferencing (Flanagan, 1985a,b;Flanagan et al, 1991;Kellerman, 1991), speech recognition (Silverman, 1987;Che, Rahim & Flanagan, 1992;Che et al, 1994;Giuliani, Omologo & Svaizer, 1994), speaker identification (Lin, Jan & Flanagan, 1994), speech acquisition in an automobile environment (Grenier, 1992;Oh, Viswanathan & Papamichalis, 1992), sound capture in reverberant enclosures (Flanagan, Surendran & Jan, 1993;Adugna, 1994;Jan, Svaizer & Flanagan, 1995), largeroom recording-conferencing (Flanagan et al, 1985), acoustic surveillance (Omologo & Svaizer, 1993, 1994, and hearing aid devices (Greenberg & Zurek, 1992). …”

A practical time-delay estimator for localizing speech sources with a microphone array

“…Existing array systems have been used in a number of applications. These include teleconferencing (Flanagan, 1985;Kellerman, 1991), speech recognition (Silverman, 1987;Che, Lin, Pearson, deVries & Flanagan, 1994), speaker identification (Lin, Jan & Flanagan, 1994), speech acquisition in an automobile environment (Grenier, 1992;Oh, Viswanathan & Papamichalis, 1992), large-room recording-conferencing (Flanagan, Johnson, Zahn & Elko, 1985), and hearing aid devices (Greenberg & Zurek, 1992). These systems also have the potential to be beneficial in several other environments, e.g.…”