Robust speech recognition by extracting invariant features

A long-term goal of machine learning is providing techniques to overcome the unwanted variations and noise to improve the recognition accuracy. In this study we used a model based on deep architectures to provide better representations of Input. In order to increase the capacity of recurrent Neural Network, we present a model to share common features across data. Also by using this method the extracted components will be invariant to speaker variations. We compared the performance of this method with our previous research. The results show that the proposed model is more useful in removing noise and unwanted variability.