Speech Enhancement Techniques for Noisy Speech in Real World Environments

Abstract

Communication between computer and human has become increasingly popular in today world. Investigation of human emotion importance is also growing in several domains. But under real world condition, speech signal is often, corrupted with several noise types and the accuracy of recognition is degraded from these noisy signal. Therefore this paper focuses on the speech enhancement techniques to develop emotion recognition system for the noisy signal in the real world environment. The various popular enhancement techniques are analyzed by adding the background noise to the clean signal using various SNR. To test the accuracy of the system, the widely used MFCC signal features are against with the SVM classifier. Results after enhancing were compared to that noisy signal and that clean signal to measure the system performance. The experimental results show the best performance algorithm and all enhancement algorithms improve the emotion recognition system performance under various SNRs level of real world background noise.