Siddalingappa, Rashmi ORCID: https://orcid.org/0000-0001-9786-8436, Hanumanthappa, M. and Gopala, B. (2018) Training Based Noise Removal Technique for a Speech-to-Text Representation Model. Journal of Physics: Conference Series.

Preview

Text Rashmi_2018_J._Phys.__Conf._Ser._1142_012019.pdf - Published Version Available under License Creative Commons Attribution. Download (1MB) | Preview

Abstract

The accomplishments of a Speech Recognition Process is significantly deteriorated by the presence of an unwanted speech signal called noise entity. This entity is present in the primary audio source. During the Speech – Recognition process, the presence of noise in the original audio signal adversely impact the output generated. Therefore, noise must be removed before performing any functions on the speech signal. With such observation of noise, it becomes essential to apply a unique procedure that perforce the noise without causing any distortion to the original audio. This research paper presents a novel approach to de-noise the given input audio signal based on the training method. Further, the paper explains the architecture adopted for Training Based Noise Removal Technique (TBNRT), steps of noise removal process, and the evaluation of the results obtained by the proposed procedure. The SNR values of the input are compared with the SNR values of the audio signal after applying the proposed TBNRT. Improvements in the SNR values were observed after the application of the proposed method. The obtained results were compared with the existing techniques and the proposed TBNRT gave promising results.

Item Type:	Article
Status:	Published
DOI:	10.1088/1742-6596/1142/1/011001
Subjects:	Q Science > Q Science (General)
School/Department:	London Campus
URI:	https://ray.yorksj.ac.uk/id/eprint/12882

University Staff: Request a correction | RaY Editors: Update this record