Application of subspace-based CNN speech noise reduction model in medical field

Xian Fu; Sisi Dong; Weiqi Zhou; Zhuzhu Zhang; Ruiwen Ye; Ting Ren

doi:10.1117/12.2680947

8 June 2023 Application of subspace-based CNN speech noise reduction model in medical field

Xian Fu, Sisi Dong, Weiqi Zhou, Zhuzhu Zhang, Ruiwen Ye, Ting Ren

Proceedings Volume 12707, International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2023); 127074J (2023) https://doi.org/10.1117/12.2680947
Event: International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2023), 2023, Changsha, China

Abstract

In order to enhance the noise reduction capability and recognition efficiency of speech recognition technology applied in medical devices, this paper adopts an improved subspace CNN based noise reduction model, adding a subspace projection module with two orthogonal self-attentive mechanisms to the original model, inputting the feature vectors extracted from the convolutional layer to the self-attentive layer, and the two self-attentive layers learn and project orthogonally to each other to obtain the noise embedding vector and speech embedding vectors to extract cleaner speech. In this paper, the improved subspace CNN noise reduction model is compared with the CNN noise reduction model by comparing the distortion degree, segmental signal-to-noise ratio and quality at different signal-to-noise ratios to demonstrate the reliability of the model. The results show that the overall distortion level of the model is reduced by about 7.27%, the overall segmental SNR performance is increased by about 15.27% and the quality is improved by about 8.7%, which is a certain improvement in performance compared with the traditional CNN noise reduction model.

Citation Download Citation

Xian Fu, Sisi Dong, Weiqi Zhou, Zhuzhu Zhang, Ruiwen Ye, and Ting Ren "Application of subspace-based CNN speech noise reduction model in medical field", Proc. SPIE 12707, International Conference on Image, Signal Processing, and Pattern Recognition (ISPP 2023), 127074J (8 June 2023); https://doi.org/10.1117/12.2680947

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available