Range Data Acquistion and Shape Feature Extraction
Dong-Uk Cho, Ji-Yeong Kim, Boo-Ho Lee · The Journal of the Acoustical Society of Korea · 1992
A new pitch detection algorithm is proposed. It takes advantage of the fact that if the phases of the fundamental and its harmonics are synchronized, the superimposed waveform shows peaks at the same peak position of the fundamental. We set the phase of the Fourier-transformed speech signal to zero, effectively synchronizing the phases of all harmonics. The algorithm gives robust performance even in 0 dB SNR environment, with a gross error rate of 3. 63%. The gross error for clean speech is only 0.18%. It also exhibits good pitch resolution since the decision logic works in the time domain. Overall experimental results indicate that the proposed algorithm is quite effective for pitch detection.