Detection and Reduction of High Frequency Non-stationary Vehicular Engine Noise Using Single Microphone

In mobile speech communications, because of the noise interfering with speech at one end, the intelligibility of speech degrades at the other end. In this paper, we focus on suppression of noise produced by vehicular and automobile mechanical engines, in whose presence the intelligibility of the speech deteriorates, when transmitted to the other end (or) while recording. We propose a method using single microphone (channel) to detect and reduce high frequency non-stationary vehicular engine noise to improve the quality of speech considerably. Simulation results demonstrate that the proposed method reduces the vehicular engine noise better than conventional noise reduction techniques. The parameters of the proposed algorithm are tuned and adapted such that optimal trade-off is achieved between noise reduction and speech retention which is also proved experimentally.