Performance evaluation of front end speech enhancement techniques

In this paper the performance of speech enhancement algorithms was evaluated and compared in terms of improvement in Signal-to-Noise Ratio (SNR) and speech quality. Clean and noisy version of one Continuous Hindi speech sentence was used for performing the experiments. Noisy version is obtained by mixing F16 noise, operations room noise and machine gun noise to clean speech signal at different SNR levels. The study of this paper leads to the utility of speech enhancement algorithms for different noises based on their characteristics.