American Journal of Circuits, Systems and Signal Processing Vol. 1, No. 2, 2015, pp. 47-55 http://www.aiscience.org/journal/ajcssp * Corresponding author E-mail address: compeasywalus2@yahoo.com (I. A. Alimi) Communication Systems Noise Reduction Based on Adaptive Spectral Subtraction Method Isiaka A. Alimi * , Tusin D. Ebinowen Department of Electrical and Electronics Engineering, School of Engineering and Engineering Technology, Federal University of Technology, Akure, Nigeria Abstract The spectral subtraction (SS) method is a well-known signal enhancement technique that reduces the effect of noise in a noisy signal in order to improve the signal quality. The SS works on the principle that noise spectrum estimate over the entire speech spectrum can be subtracted from the noisy signal. However, noise does not affect the speech signal uniformly over the entire spectrum at different frequency bands. Therefore, most implementations of the basic technique lead to anomaly known as “musical” tones artifacts in the enhanced signal. The abnormality can then be perceived as residual noise and speech distortion in the resulting signal. In this paper, we propose a multi-band spectral subtraction (MBSS) method using novel noise element suppression (NES). The proposed scheme gives comparatively better performance and the computation required is minimal. Furthermore, simulation results show that the proposed algorithm removes noise without removing the relatively low amplitude signal over the entire speech spectrum. Keywords Speech Enhancement, Musical Noise, Spectral Subtraction, Noise Element Suppression, Multi-Band, Sub-Band Received: May 2, 2015 / Accepted: May 17, 2015 / Published online: June 18, 2015 @ 2015 The Authors. Published by American Institute of Science. This Open Access article is under the CC BY-NC license. http://creativecommons.org/licenses/by-nc/4.0/ 1. Introduction In recent years, voice communication technologies have been rapidly growing since the advent of mobile telephony systems in the commercial market. As the technologies evolve, new services are also introduced like teleconference systems and voice over internet protocol (VOIP). Noise suppression techniques are essential for the systems to operate efficiently and effectively because, the presence of noise often result in erroneous and unreliable communication systems [1]. Consequently, a method that can suppress the noise while maintaining the required sound quality is essential [2]. Boll proposed SS technique for suppressing the effect of noise acoustically added to the speech signals [3]. The approach is popular because of its simplicity and flexibility in concept and effectiveness in enhancing speech degraded by additive noise [4]. To implement SS, spectral magnitude of the received noisy signal is estimated and that of the noise spectrum is estimated from regions that are analyzed as “noise-only” using voice activity detection (VAD). The magnitude spectrum of noise is then subtracted from that of the noisy signal. The approach works under the assumption that noise signal is uncorrelated and remains relatively constant prior to and during voice activity. This implementation noticeably gives quality speech signal but generates musical noise through linear subtraction of noise across the entire speech spectrum [5], [6]. Current researches focus on a nonlinear subtraction process which is justified by the variation of signal-to-noise ratio across the speech spectrum. Also, the nonlinear approach shows that noise signal does not affect the speech signal uniformly over the whole spectrum because certain frequencies are affected more adversely than others [2], [7]. To prevent the variation of signal-to-noise ratio across the