During the acquisition of the speech signal by the non-contact Speech Sensor (SS), the signal is degraded by severe colored noises which are non-linear and non-uniform in nature. Therefore, in this study, a new approach for advancing the speech enhancement technique is proposed to suppress these noises from the acquired speech signal. This technique is based on adaptive thresholding, which uses Mean Median Deviation(MMD) method to determine adaptive threshold points and the Empirical Mode Decomposition (EMD) method. This algorithm has been validated by simulation data, and the results of the proposed algorithm have been compared with other existing enhancement algorithms. Spectrograms and Signal to Noise Ratio (SNR)comparison are used to analyze the quality of the enhanced speech signal. From the results, we demonstrate that the proposed algorithm offers better speech enhancement than previously existing speech enhancement techniques.
Please cite the following BibTex:
@inproceedings{dubey2019speech,
title={Speech Enhancement using Adaptive Mean Median Deviation and EMD Technique},
author={Dubey, Shikha and Singh, Ashish Kumar and Singh, Manoj Kumar},
booktitle={2019 IEEE International Conference on Signals and Systems (ICSigSys)},
pages={34--39},
year={2019},
organization={IEEE}
}
If you find the paper and this repository helpful, please consider citing our paper . Thank you!