A Novel Magnitude-Squared Spectrum Cost Function for Speech Enhancement
AbstractIn Bayesian approaches for speech enhancement, the enhanced speech is estimated by minimizing the Bayes risk. In detail, an estimate of the clean speech is derived by minimizing the expectation of a cost function. Various estimators have been derived by the classic cost function, squared-error cost function, and “hit-or-miss” function. However, absolute error function was paid less attention. In this paper, we consider a magnitude-squared spectrum (MSS) motivated estimator for speech enhancement based on statistics and Bayesian cost function in the frequency domain. Specifically, we derive a novel estimator of which the cost function is the absolute error distortion measure of the MSS. By studying experimental results with NOIZEUS database, we find that the performance of the proposed scheme can achieve a significant noise reduction and a better speech quality as compared to minimum mean-squared error (MMSE) estimator of the MSS. Ill. 1, bibl. 14, tabl. 2 (in English; abstracts in English and Lithuanian)
How to Cite
The copyright for the paper in this journal is retained by the author(s) with the first publication right granted to the journal. The authors agree to the Creative Commons Attribution 4.0 (CC BY 4.0) agreement under which the paper in the Journal is licensed.
By virtue of their appearance in this open access journal, papers are free to use with proper attribution in educational and other non-commercial settings with an acknowledgement of the initial publication in the journal.