Frame Decorrelation for Noise-Robust Speech Recognition

Abstract

The authors propose a frame decorrelation method to cope with background noise in speech recognition. Since noise is modelled as a stationary perturbation in most cases, it is effective to reduce slow-varying components. One example of using this principle is the highpass scheme. The proposed method has the same property as the highpass scheme. It transforms feature vector sequences into decorrelated sequences and enhances transition regions. Simulation results show that this method is effective for speech with significant noise, and works better than other highpass methods.