Research on the Application of Speech Recognition Technology Based on Transformer Model

Speech recognition technology has developed from the 1950s to the present, evolving from template matching methods to Hidden Markov Model (HMM) statistical methods, then to machine learning techniques, and finally to the current use of Transformer technology for speech recognition tasks. However, the Transformer model has not yet been widely adopted in the field of speech recognition. This paper explores the characteristics of Transformer model, combines it with the characteristics of speech recognition tasks, analyzes the challenges associated with using Transformer model for these tasks, and provides suggestions for directions of future research, so as to facilitate the application of Transformer models in speech recognition. The paper finds that the reasons for the limited application of Transformer models in speech recognition tasks mainly include their numerous parameters, complex structure, and high computational costs, which have prevented their extensive use in this field. In the future, efforts should focus on enhancing model compression and lightweight design, and improving the attention mechanism to boost the applicability of Transformer models in speech recognition.