Research Article
Anti Noise Speech Recognition Based on Deep Learning in Wireless Communication Networks
@INPROCEEDINGS{10.1007/978-3-031-50543-0_11, author={Yanning Zhang and Lei Ma and Hui Du and Jingyu Li}, title={Anti Noise Speech Recognition Based on Deep Learning in Wireless Communication Networks}, proceedings={Advanced Hybrid Information Processing. 7th EAI International Conference, ADHIP 2023, Harbin, China, September 22-24, 2023, Proceedings, Part I}, proceedings_a={ADHIP}, year={2024}, month={3}, keywords={Deep Learning Improve EMD Hidden Markov Model Anti Noise Speech Recognition}, doi={10.1007/978-3-031-50543-0_11} }- Yanning Zhang
Lei Ma
Hui Du
Jingyu Li
Year: 2024
Anti Noise Speech Recognition Based on Deep Learning in Wireless Communication Networks
ADHIP
Springer
DOI: 10.1007/978-3-031-50543-0_11
Abstract
As a new high-tech industry, the application of speech recognition technology is becoming more and more competitive, with a wide range of application fields and application prospects, and has far-reaching significance for the development of science and technology. The communication environment of wireless communication network will bring various types of noise to speech, so an anti noise speech recognition method based on deep learning of wireless communication network is designed to achieve anti noise speech recognition in this environment. The voice signal of wireless communication network is preprocessed by anti aliasing filtering, analog-to-digital conversion, pre emphasis, framing and windowing, endpoint detection, etc. A series of denoising processes are implemented for the voice signal of wireless communication network, and different speech preprocessing methods are adopted for different characteristics of noise. A speech signal feature extraction method based on improved EMD is designed and implemented. The speech recognition model is designed based on the regression neural network in deep learning, and the anti noise speech recognition of wireless communication network is realized. Test results show that the lowest word error rate of this method is 0.156, and the word error rate is also low.
