Research on Mongolian Speech Recognition Based on the Self-supervised Model

As an important branch of artificial intelligence application, speech recognition is becoming more and more widely used in daily life. In the field of speech recognition of Mongolian language, its unique language structure and pronunciation system have brought new challenges to the development of speech recognition technology. As the self-supervised speech representation model is pre-trained using only audio data, without paired text labels, it achieves excellent performance on a variety of downstream tasks. This paper details the structure and working principle of the Mockingjay model and why it is suitable to handle the problem of Mongolian speech recognition. In order to improve the performance and robustness of the model, we propose a speech processing method of Mockingjay model for features extraction of Mongolian speech data and fuse it with MFCC (Mel frequency inversion coefficient). Finally, the effectiveness of the proposed method is verified by entally.