Analysis of the influence of signal compression techniques for voice disorder detection through filter-banked based features

Abstract

This paper compares the results of using compressed voice signals versus uncompressed speech signals to automatically detect voice abnormalities. Coding techniques and voice compression used in this study are the same as those used by default in the fixed, mobile and ip telephony systems, and techniques of characterization and classification used are also among the most used for detecting automatic speech abnormalities. The results obtained indicate that it is possible to use compressed voice signals for automatic detection of vocal pathologies without compromising the success rate in the diagnosis, which would make the implementation of automatic remote diagnosis of vocal pathologies possible.