Title: Competitive crow search algorithm-based hierarchical attention network for dysarthric speech recognition

Authors: Bhuvaneshwari Jolad; Rajashri Khanai

Addresses: Department of Electronics and Communication Engineering, KLE Dr. M.S. Sheshgiri College of Engineering and Technology (MSSCET), Belagavi, Karnataka, India; Department of Electronics and Telecommunication Engineering, Dr. D.Y. Patil Institute of Technology, Pimpri, Pune, Maharashtra, India ' Department of Electronics and Communication Engineering, KLE Dr. M.S. Sheshgiri College of Engineering and Technology (MSSCET), Belagavi, Karnataka, India

Abstract: The common difficulty of speech recognition is articulation deficiency produced by an athetoid, which is a kind of cerebral palsy. In this paper, the effectual dysarthria speech recognition approach is introduced using the developed Competitive Crow Search Algorithm-based Hierarchical Attention Network (CCSA-based HAN). Here, the spectral subtraction method is utilised for removing unwanted noises. Then, the specific features are extracted, and then to improve the performance the data augmentation is done. The data augmentation process is performed by adding various noises, like street noise, train noise and party crowd noise to the input signal. In addition, the HAN classifier is employed for recognising dysarthric speech. Here, CCSA is devised for obtaining effective recognition output, which is designed by incorporating Competitive Swarm Optimiser (CSO) and Crow Search Algorithm (CSA). The developed dysarthria speech recognition approach outperforms other existing methods with accuracy of 0.9141, sensitivity of 0.9208 and specificity of 0.9172.

Keywords: hierarchical attention network; dysarthric speech recognition; competitive swarm optimiser; crow search algorithm.

DOI: 10.1504/IJWMC.2023.135384

International Journal of Wireless and Mobile Computing, 2023 Vol.25 No.4, pp.340 - 352

Received: 08 Sep 2021
Received in revised form: 17 Aug 2022
Accepted: 08 Sep 2022
Published online: 08 Dec 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article