Implementation of real-time AMDF pitch-detection for voice gender normalisation

Abstract

Traditionally the interest in voice gender conversion was of a more theoretical nature rather than founded in real-life applications. However, with the increase in mobile communication and the resulting limitation in transmission bandwidth new approaches to minimising data rates have to be developed. Here voice gender normalisation (VGN) presents a novel method of achieving higher compression rates by using the VGN algorithm to remove all gender specific components of a speech signal and thus leaving only the information content to be transmitted. A second application for VGN is in the field of speech controlled systems, where current speech recognition algorithms have to deal with the voice characteristics of a speaker as well as the information content. Here again the use of VGN can remove the speakers voice characteristics leaving only the pure information. Therefore, such a system would be capable of achieving much higher recognition rates while being independent of the speaker. This paper presents the theory of a gender removal system based on VGN and furthermore, outlines an efficient real-time hardware implementation for use in portable communications equipment.