A study of pitch detection techniques

Abstract

The main aim of this theses was to investigate short-time pitch detection algorithms. Many such algorithms exist and only two of these were studied in detail. These were the SIFT algorithm, and another algorithm employing autocorrelation . and centre-clipping techniques. Both algorithms work in the time domain, although the SIFT algorithm utilises both time and frequency domain properties of speech signals. The algorithms were implemented in software using Turbo C and tested on speech samples. Use was also made of the hypersignal software found in the university, to check the results obtained.