Pitch estimation using music algorithm based on the sinusoidal speech model

In this paper we present a new method of pitch estimation in speech signals. The existing methods in pitch estimation do not provide an accurate estimate of the pitch value, in presence of sinusoidal noise. Thus these methods have large errors in presence of sinusoidal noise and some other noise characteristics. The method we presented in this paper is based upon the MUSIC algorithm to estimate the pitch frequency. MUSIC algorithm has been widely used to estimate the frequency and angle of arrivals for narrow(wide) band singals. We note that the algorithm presented here accurately estimates the pitch frequency, in the presence of sinusoidal noise and like the other methods in the presence of Gaussian noise.