A model of auditory perception as front end for automatic speech recognition

Abstract

A front end for automatic speech recognizers is proposed and evaluated which is based on a quantitative model of the "effective" peripheral auditory processing. The model simulates both spectral and temporal properties of sound processing in the auditory system which were found in psychoacoustical and physiological experiments. The robustness of the auditory-based representation of speech was evaluated in speaker-independent, isolated word recognition experiments in different types of additive noise. The results show a higher robustness of the auditory front end in noise, compared to common mel-scale cepstral feature extraction. In a second set of experiments, different processing stages of the auditory front end were modified to study their contribution to robust speech signal representation in detail. The adaptive compression stage which enhances temporal changes of the input signal appeared to be the most important processing stage towards robust speech representation in noise. Low-pass filtering of the fast fluctuating envelope in each frequency band further reduces the influence of noise in the auditory-based representation of speech.

Robust Distributed Speech Recognition Using Auditory Modelling

“…In that study, there was no pre-processing or enhancement of the speech utterances. The front-ends investigated were perceptual linear prediction (PLP) proposed by Hermansky [17], the PEMO algorithm proposed by Tchorz and Kollmeier [18], and the front-end processor proposed by Li et al [14]. For the task of connected digit recognition using the Aurora 2 database, the front-end proposed by Li et al gave the best overall recognition results of all the auditory models examined, and with an overall reduction in recognition error compared to the ETSI basic front-end [6] which was used as a baseline for comparison.…”

A computational model of mechanisms controlling experience-dependent reorganization of representational maps in auditory cortex

“…This connectionist model has a highly flexible architecture that can be easily customized to model processing in sensory cortices (see, e.g., Erwin et al, 1995;Palakal et al, 1995;Ritter et al, 1992;Sirosh & Miikkulainen, 1997;Swindale & Bauer, 1998). Signalprocessing models of auditory representations have advanced considerably over the past decade (see, e.g., Meyer-Base & Scheich, 1995;Pitton, Wang, & Juang, 1996;Robert & Eriksson, 1999;Tchorz & Kollmeier, 1999;K. Wang & Shamma, 1995a).…”

Minimum variance modulation filter for robust speech recognition

“…[4,5]), a number of feature extraction methods have been proposed in recent years that exploit temporal information. These systems typically provide a recognition accuracy that exceeds that obtained using MFCC or PLP features in the presence of noise and other adverse conditions [6,7,8], especially if they are combined with a traditional recognition system in some fashion.…”