• Naik D C
• Sreenivasa Murthy A

Over the last 50 years or so researches/engineers have proposed a number of speech enhancement algorithms. The goal of speech enhancement algorithm is to obtain the clean speech to a level that is, appreciated by human listeners be impaired or normal and that improves the performance of speech-activated machines. Spectral subtraction is probably the first approach proposed in this direction. It is so attractive in terms of both comprehension and implementation that even after 40 years of its existence, researches looking at modifications to the basic approach to get a better signal in the sense of improved intelligibility and quality. In most spectral subtraction based speech enhancement approach, the estimated noise spectrum is assumed to affect the entire short time speech spectrum uniformly. However for most of the real world noise scenario, these assumptions turn out to be invalid to alleviate this problem, the spectrum of each frame is divided into a number of sub-bands and weighted noise estimates are subtracted from each of these sub-bands. The noise estimates are obtained from the sub-band specific SNRs. The multi-band spectral subtraction alleviates this problem by subtracting the weighted noise estimates from the non-overlapping bands of the noisy speech spectrum. In this research activity, we have used multi-band spectral subtraction in magnitude spectral domain and tested for babble noise, random noise, car noise and helicopter noise using both subjective and objective measures. The result is compared with those obtained by power spectral domain. It is observed that the method results in reduced background noise and improved speech quality for babble noise and random noise.

spectral subtraction, multi-band spectral subtraction, subjective and objective performance measures, spectrograms

"A Study of Multiband Spectral Subtraction for Speech Enhancement in Magnitude and Power Spectral Domains", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.6, Issue 6, page no.257-262, June-2019, Available :http://www.jetir.org/papers/JETIR1906470.pdf

"A Study of Multiband Spectral Subtraction for Speech Enhancement in Magnitude and Power Spectral Domains", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.6, Issue 6, page no. pp257-262, June-2019, Available at : http://www.jetir.org/papers/JETIR1906470.pdf