Application of multiwavelet filter to pitch period detection
Wang Xiu-xin, Guoyu Xu, Weiping Hu, Dongdong Liang, Xiaochun Lu, Qiang Wang · Computer Engineering and Applications Journal · 2007
Pitch period is one of the important characteristic factors with the voice signals.It is often used to detect pathological voices in the clinic.Therefore first the biorthogonal multiwavelet filter was constructed.Multiwavelet filter with autocorrelation was proposed to detect pitch period in speech signals.The short sentence including six Chinese phonetic vowels was introduced.Then the low-frequency signal was separated from the speech signals with the multiwavelet filter.Finally,the pitch period was detected with autocorrelation.The result indicates the method has high correctness,high accuracy,and high anti-noise in tracking pitch period.It is more effective than autocorrelation and single wavelet with autocorrelation under the circumstances of the various noises.Especially in the circumstances of the high noise it has clear anti-noise ability.It isn't affected by the non-stationary characteristics of the speech signals.It can be used to detect the pitch period from pathological voices effectively.