Research on Speech Recognition Algorithm Based on Embedded System Architecture
Wei Li, Hongyun Bin · 2023
In order to improve the application value of speech recognition algorithm in embedded system, the speech recognition algorithm is optimized and designed using the embedded system structure. Under the embedded system structure model, the hardware equipment is used to collect the voice signal, and the initial voice signal is preprocessed through filtering, windowing and framing, endpoint detection and other steps. From the two dimensions of cepstrum domain and time domain, the speech signal features are extracted, and the speech recognition results are obtained by matching with the language template features. Compared with the traditional recognition algorithm, it is concluded that the speech recognition error rate of the optimized design algorithm is reduced by about 1.21% and 0.87% respectively, and the recognition dropout rate is reduced by 1.38% and 0.66% respectively in the environment with and without noise, which proves that the optimized design algorithm has higher recognition performance.