The acoustic features based on pitch detection after process analysis

In the field of speech signal processing, a pitch detection algorithm (PDA) is commonly used to estimate pitch or fundamental frequency. If the given speech signal is clean, the algorithm can achieve better detection result. However, in general, the speech signal would inevitably influenced by the background noise, the detection algorithm may not work well and the detected pitches may deviated from the correct position. In this paper we propose a new approach for improving the accuracy of pitch detection. This approach uses a median filter to remove the outliers in the results produced by short-time autocorrelation or cepstrum method. The conducted experiments show that the proposed approach works well.