Speech Noise Reduction Based on Frequency Domain Adaptive Line Enhancer

We propose the speech noise reduction system based on the frequency domain adaptive line enhancer. The adaptive line enhancer (ALE) is effective to extract sinusoidal signals blurred by a broadband noise. The ALE utilizes only one microphone; therefore, it is suitable for miniaturizing portable electronic devices. Especially, our proposed system adopts the frequency domain adaptive filter by using the modified DFT in order to obtain faster convergence than in time domain. In addition, a desired signal is generated by delaying an input signal by the de-correlation parameter in the ALE, which makes a noise in the desired signal de-correlated with that in the input one. In the proposed system, we introduce an individual de-correlation parameter in each frequency domain signal and adjust it to guarantee the correlation of the speech element. Noise reduction performance is examined through computer simulations using practical speeches.