Pitch Estimation Algorithm for Narrowband Speech Signal using Phase Differences between Harmonics

This paper proposes a pitch estimation algorithm for a narrowband speech signal using phase differences between har-monics. A narrowband speech signal has an incomplete harmonic structure due to bandwidth limitation, which degrades the pitch estimation accuracy. In this paper, we focus on the fact that phase differences between harmonics are constant by approximating a speech signal based on a sinusoidal model. Since the narrowband speech signal has a partial harmonic structure, the proposed method selects the pitch such that the phase differences between harmonics are constant on the narrow bandwidth. Also, we take a frame additive average method for the phase differences between harmonics to improve the robustness against noise. Experimental results show that the proposed method estimates the pitch from the narrowband speech signal under noisy environments more accurately than the traditional methods.