A novel pitch detection algorithm for noisy speech signal based on Radon transform and multi-frame correlation

In this article we propose a novel fundamental frequency detection algorithm for noisy speech signals. The algorithm combines Radon transform and multi-frame signals correlation to extract the fundamental frequency, that is, pitch period from voiced frames in degraded speech signals. Two publicly available datasets, the CSTR and TIMIT datasets, were used to evaluate the performance of the algorithm and other state-of-the-art pitch detection algorithms under various additive daily environmental noises conditions and multiple signal-to-noise ratios. As far as the Gross Pitch Error and Mean Absolute Error metrics are concerned, the results demonstrate that the proposed method achieves better results among all the algorithms in general.