A Statistical Approach for Speech Enhancement in Cognitive Radio Network

It is the inherent nature, that speech signal is noisy due to involvement of unvoiced component. While it passes through wireless channel the signal corrupts even more due to channel noise. Wireless network may be either licensed or unlicensed but the signal transmission and reception is an important fact. To avoid the congestion in network, the unlicensed spectrum is well utilized through one of the method named as cognitive radio network. In this paper, authors have approached to enhance speech signal at the user end so that the desire of getting clean signal can be fulfilled. The signal is normalized with low pass filter and transmitted through multicarrier modulation system using OFDM technique. Maximum likelihood (ML) technique has been utilized with cooperative spectrum sensing environment to enhance the signal communicated through cognitive radio channel. As channel noise varies 10 to 40 dB of the additive Gaussian noise, it has been verified through simulation. However, the result is shown for 10 dB noise with twenty user based cognitive radio network environment. In first stage the well-known cooperative spectrum sensing model has been implemented. Next to it the statistical approach is utilized to enhance the signal. Both short and long speech signals are verified. The result is found from visual inspection as well as the complementary ROC shows the probability of false alarm and probability of mixed detection.