Neural-Network-based Pitch-lag Estimation in Long-term Linear Prediction

This paper presents a novel method for neural-network-based pitch-lag estimation implemented within the framework of linear prediction. In comparison to the standard autocorrelation pitch-lag estimation technique, the proposed approach exploits the capability of AI algorithm, namely a re-current neural network, to enhance the accuracy of long-term linear prediction by means of efficient pitch-lag identification. The proposed approach not only showcases the potential of AI to outperform traditional methods in signal processing but also opens new ways for refining the analysis of pitch-related phenomena in diverse applications.