A New Data Driven Method for Robust Speech Recognition

The conventional view on the problem of robustness in speech recognition is that performance degradation in ASR systems is due to mismatch between training and test conditions. If problem of robustness in ASR systems were considered as a mismatch between the training and testing conditions the solution would be to find a way to reduce it. Common approaches are: Data-Driven methods such as speech signal enhancement and using robust features and model-based methods that alliterate or adapt model of speech signal. In this paper, we review some of data-driven and model-based methods and implement some of datadriven methods such as: spectral subtraction, cepstral mean normalization, cepstral mean and variance normalization, mapping noisy space to clean space and SNR-depndent cepstral normalization. We suggest a new data-driven method for robust speech recognition using neural networks and compare it to other methods. Key-Words: Robust speech recognition, cepstral normalization, neural networks