Electrical & Electronics Engineering

Permanent URI for this collectionhttp://197.211.34.35:4000/handle/123456789/65

Electrical & Electronics Engineering

Browse

Subject: search.filters.subject.Non-parametric

Search Results

Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Non-intrusive noise reduction in GSM voice signal using non-parametric modeling technique
    (International Engineering Conference (IEC), School of Engineering and Engineering Technology, FUT Minna, 2015-06-15) Gbadamosi Safiu Abiodun,; A. M. Aibinu,; O.C.Ugweje,; A. J Onumanyi ,; E. N Onwuka,; M. Aderinola
    Noise degrades the quality and intelligibility of speech. It impedes speech clarity, coding, recognition and speaker identification. To mitigate noise effect and improve speech quality, we propose Non-parametric modeling technique along with a Non-intrusive signal denoising system based on short time Fourier transforms. This paper aims to establish only the idea behind our proposed algorithm, however, we present argument to justify that our results will reduce an end to end acoustic background noise; improve quality of speech for both the speaker and the listener and eventually increase throughput. Ultimately, users’ will be able to call and receive calls in a noisy environment while enjoying clarity of voice.
  • No Thumbnail Available
    Item
    Development of Non-Parametric Noise Reduction Algorithm for GSM Voice Signal
    (ABUAD Journal of Engineering Research and Development (AJERD), 2018-12-31) Gbadamosi, Safiu Abiodun,; AIBINU, Musa Abiodun ,; ONUMANYI, Adeiza James
    Speech enhancement in Global System for Mobile communication (GSM) is an area of engineering that study different kinds of techniques used in enhancing GSM voice signals. The presence of noise in GSM degrades the quality and intelligible of speech which impedes speaker identification and sound clarity. In this paper, non-parametric noise reduction algorithm which incorporates an adaptive threshold technique is proposed to estimate the adaptive threshold value as a function of first and second order statistics of the voice signal. It uses the cumulative value of minimum mean and maximum standard deviation value and minimum (mean and standard deviation) to minimize the effect of impairments introduced by background noise and GSM channels. The algorithm was implemented in MATLAB environment. The results obtained indicate correlation coefficients of 53.93% and 45.52% for maximum and minimum threshold value at 2.95 standard deviation of noise and 77.93% and 75.83% for maximum and minimum threshold value at 0.25 standard deviation of noise, respectively. Evaluation of the proposed algorithm was performed on real noisy voice signal and a correlation of 92.15% and 89.24% was achieved for both maximum and minimum threshold values with mean square error of 0.0011% and 0.00033%, respectively. These results have proven the efficiency of the proposed algorithm. The threshold values have satisfied perfect noise reduction when the mean and standard deviation values are selected properly