Development of spectral subtraction algorithm for enhancement of noisy speech signal of electricity generator


Speech enhancement entails a process of reducing noise and distortions by increasing the quality and intelligibility of a speech signal. This paper presents evaluation of spectral subtraction algorithm for noisy speech (samples taken in an environment where electricity generator is operated) without losing any part of the speech signal in terms of quality, quantity and without much computational and time complexity enhancement at different signal to noise ratios (SNR). Spectral subtraction was carried out on noisy speech samples at different SNR. The Noise removal algorithm was implemented using Matlab software. The corresponding spectrum was computed using the DFT (Discrete Fourier Transform) which removes the noise from the noisy speech and the corresponding spectrum was reconstructed in the time domain using the Inverse Discrete Fourier Transform (IDFT). The algorithms performance was evaluated by varying the Signal to Noise Ratio (SNR). The result indicates the optimal SNR values for electric generator noisy Speech Samples at -5dB, 5dB, 10Db, 15Db and 20dB. The spectral subtraction algorithms perform excellently in SNR range of -5.0000dB to 17.0500dB without any loss of part of the speech signal.


No have any comment !
Leave a Comment