Discriminative Auditory Features for Robust Speech Recognition

Recently, Li et al. proposed a new auditory feature for robust speech recognition in noise environments. The new feature was derived by mimicking closely the function of human auditory pro-cess. Several lters were used to model the outer ear, middle ear, and cochlea, and the initial lter parameters and shapes were ob-tained from crude psychoacoustics results, experience, or experi-ments. Although one may adjust the feature parameters by hand to get better performance, the resulting feature parameters still may not be optimal in the sense of minimal recognition errors, espe-cially for different tasks. To further improve the auditory feature, in this paper we apply discriminative training to optimize the audi-tory feature parameters with some guidance from psychoacoustic evidence but otherwise in a data-driven approach so as to mini-mize the recognition errors. One signicant contribution over sim-ilar efforts in the past, such as discriminative feature extraction, is that we make no assumption on the parametric form of the audi-tory lters. Instead, we only require the lters to be smooth and triangular-like as suggested by psychoacoustics research. Our ap-proach is evaluated on the Aurora database and achieves a word error reduction of 19.2%. 1