Research on Spanish Classroom Speech Recognition System Based on Deep Learning Algorithm

This study designs and optimizes a Spanish speech recognition system based on deep learning. By combining Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks, an efficient speech recognition model is constructed. The experiment utilizes diverse Spanish accent datasets, the ESPnet dataset, and the TTS dataset, and employs data augmentation and hyperparameter tuning techniques to enhance the model's performance. Experimental results show that the CNN-LSTM-based model outperforms the traditional HMM-GMM method in terms of word error rate (WER), accuracy, and recall. The Transformer-based model further improves recognition accuracy, demonstrating the advantages of deep learning methods in variable environments.