Robust pitch estimation using an event based adaptive Gaussian derivative filter

In the development of practical speech processing algorithms, the ability to automatically and accurately determine the pitch period in noisy environments remains a fundamental obstacle. In this paper, we propose a new pitch detection algorithm based on an iterative adaptive smoothing approach using a Gaussian Derivative filter which is the sum of a zeroth and second order Hermite function. We refer to this new algorithm as the Adaptive Gaussian Derivative Filter (AGDF). The AGDF pitch detector works under varying noise conditions, with variable pitch periods and for different speakers. We compare the performance of the AGDF method to the approach based on the Dyadic Wavelet Transform (DyWT) and the pitch prediction (PP) formulation for speech subjected to different noise conditions and signal to noise ratios (SNR). The results show that the AGDF is slightly better than the DyWT pitch detection scheme and significantly outperforms the PP approach.