Speech enhancement using transform domain techniques

Abstract

The techniques of speech enhancement are used to minimize the undesirable external noises; the speech signal get degraded due to these external noises and these noises are background noise, wind noise, and other noises. These noises are generated due to environmental sources such as vehicular noise, street noise, babble noise, etc. Therefore, the speech enhancement plays an important role to reduce unwanted noise from the speech signal. This paper shows the single channel, the speech enhancement techniques based on a spectral subtraction method (SSM) and wavelet packets method (WPM) have been used for removing the noises. The performance of enhanced speech signal has been evaluated by measuring the speech quality parameters such as cross-correlation, average absolute distortion (AAD), peak signal-to-noise ratio (PSNR), and mean square error (MSE). The speech signal recorded in male voice and the Noizeus Speech database with different speech signals has been used to test the performance of this system. The experimental result shows that the WPM is more suitable for speech enhancement.