Selective Gammatone Envelope Feature for Robust Sound Event Recognition

Abstract

SUMMARYConventional features for Automatic Speech Recognition and Sound Event Recognition such as Mel-Frequency Cepstral Coefficients (MFCCs) have been shown to perform poorly in noisy conditions. We introduce an auditory feature based on the gammatone filterbank, the Selective Gammatone Envelope Feature (SGEF), for Robust Sound Event Recognition where channel selection and the filterbank envelope is used to reduce the effect of noise for specific noise environments. In the experiments with Hidden Markov Model (HMM) recognizers, we shall show that our feature outperforms MFCCs significantly in four different noisy environments at various signal-to-noise ratios.

A Spiking Neural Network Framework for Robust Sound Classification

“…Several feature representations for acoustic signals have been proposed over the years for capturing frequency contents and temporal structures of acoustic signals (Mitrović et al, 2010 ). The most frequently used features are the Mel-Frequency Cepstral Coefficients (MFCC) (Chu et al, 2009 ) and Gammatone Cepstral Coefficients (GTCC) (Leng et al, 2012 ). Both these features mimic the human auditory system, as they are more sensitive to changes in the low-frequency components.…”

Sound event recognition in unstructured environments using spectrogram image processing

“…Further to this, recent works have also demonstrated improved performance through pitch-adaptivity [135] or selection [136] of the gammatone filterbanks. For example, in [136], filterbank channel selection is performed to adapt an SER system to changing environmental conditions. This was shown to outperform both MFCC and selective Mel-filterbank features.…”

“…For non-speech audio classification, it has been shown that GTCCs can outperform both MFCCs and MPEG-7 using both kNN and SVM classifiers [86]. Further to this, recent works have also demonstrated improved performance through pitch-adaptivity [135] or selection [136] of the gammatone filterbanks. For example, in [136], filterbank channel selection is performed to adapt an SER system to changing environmental conditions.…”

Sound Events Localization and Detection Using Bio-inspired Gammatone Filters and Temporal Convolutional Neural Networks

“…By contrast, Gammatone filter impulse responses were obtained from measures on the basilar membrane of small mammals. Moreover, applying a Gammatone filterbank to the spectrogram has shown to be more robust against ambient noise in acoustic event monitoring compared with Mel-scale filterbank representations [20], [21]. The Gammatone filterbank has also shown good performance in automatic audio captioning systems [22] and active noise control systems [23].…”