This paper describes a method for pitch extraction of musical instrument tones using digital signal processing techniques. The digitized data of musical instrument tones are divided into sequential waves of short-period frames. then Discrete Fourier Transform(DFT) is applied to them. The same operation is successively applied to each frames by one period of the fundamental waveform. The pitch variation Δω_<n-1, n> is given as a function of time sequences, Δω_<n-1, n>=(φ_n-φ_<n-1>)/ΔT. Here, φ_n and φ_<n-1> are phase of n th and (n-1)th frame, and ΔT is shift time. Using synthesized waveforms, the errors of pitch detection were estimated. The calculated result shows that the good accuracy is obtained for pitch detection except the attack transient part of musical instrument tones which has rapid amplitude variation. Finally, this method is applied to theree natural musical instrument tones, and the results show outstanding feature of pitch variation for the attack transient, the pseudo steady state and the decay part.
抄録全体を表示