A Diagnostic Tool For The Evaluation Of A Speech-Analysis Method (NOT PUBLISHED)

Most speech processing systems (e.g., speech recognition systems or speech coding systems) contain a feature-analysis stage that extracts the required information from the speech waveform. Fo For example, current recognition systems use cepstral coefficients and their time derivatives. Another technique is to use an auditory-based representation, aiming to preserve information that is perceptually most relevant. In either technique, part of the speech information that was contained in the time waveform is lost during the process of feature extraction. This study addresses the question of how to identify what part of the speech information is lost.