Auto-Associative Initialization of LSTM Neural Networks for Fundamental Frequency Detection in Noisy Speech Signals

Abstract

In this paper, we present a new approach for fundamental frequency detection in noisy speech, based on Long Short-term Memory Neural Networks (LSTM). Fundamental frequency is one of the most important parameters of human speech. Its detection is relevant in many speech signal processing areas and remains an important challenge for severely degraded signals. In previous references for speech enhancement and noise reduction tasks, LSTM has been initialized with random weights, following a back-propagation through time algorithm to adjust them. Our proposal is an alternative for a more efficient initialization, based on a supervised training using an Auto-associative network. This initialization is a better starting point for the fundamental frequency detection in noisy speech. We show the advantages of this initialization using objective measures for the parameter and the training process, with artificial noise added at different signal-to-noise levels. Results show the performance of the LSTM increases in comparison to the random initialization, and represents a significant improvement in comparison with classic algorithms of paramater detection in noisy conditions.