A secondary feature algorithm for pitch determination of speech signals and its implementation in real-time using a Motorola DSP56001 based system

The pitch of speech is a useful parameter both in its own right and as a means of improving the estimates of other parameters. For instance, formant analysis is used to estimate the shape of the vocal tract and hence determine the speech sound uttered. If a reliable pitch estimate is available, then pitch-synchronous analysis may be used to estimate the formants, yielding a better result. Although reliable pitch estimates may be obtained using a transducer known as a laryngograph, it is generally a requirement of a practical system that the processing is carried out on the speech signal itself, i.e. the electrical signal obtained using a microphone. This paper presents an algorithm for pitch determination of speech, and discusses its implementation using digital signal processing technology.<>