Voice Conversion Algorithm

Abstract

Recently, a lot of work has been done in the speech technology. The main concentration being on Text-to-speech and automatic speech recognition techniques, voice conversion is yet an undeveloped and naïve field in Speech Technology and a lot of contribution from speech researchers is expected in upcoming days. In this paper an approach for static voice conversion is discussed. Static speech parameters are the parameters over which speaker has least control such as vocal tract structure, natural pitch of speech etc. Here, two main parameters are considered Vocal Tract Structure and Pitch. For conversion process speech is resolved in two components, excitation component and filtered component using a Linear Predictive coding [LPC] based source-filter. The pitch contour is determined by an autocorrelation. The excitation component is generated using a set of signal generators generating the determined pitch and are driven by voicing detection. Filter coefficients are modified to approach target speaker coefficients for voiced segments and for unvoiced segments filter coefficients of source are used straightaway.