Recognition algorithm of piano playing music notes based on improved hidden Markov model

The recognition process of piano playing music notes is affected by low-frequency components, resulting in suboptimal features that affect recognition accuracy. Therefore, an improved hidden Markov model based recognition algorithm for piano playing music notes is proposed. The audio signal of piano playing music in real life is collected, to pre emphasize and pre process the audio signal after obtaining it, and improve the quality of audio signal of piano playing music; Extract signal features and emotional features from piano playing music audio signals in sequence, fuse the two features to form a piano playing music signal feature set, input the fused features into a hidden Markov model, and calculate the output probability score of the features on HMM through time series modeling. Based on the score, select the optimal features. And input the obtained optimal features into the GA-RBF neural network for learning. Reduce the impact of low-frequency components on recognition results to obtain the best piano playing music note recognition results. Experimental verification shows that this method can effectively improve the quality of collected audio signals, recognize corresponding notes during different piano playing periods, and recognize corresponding notes at different frequencies. And in the process of note recognition, only 0.3 ms is needed. The proposed method optimizes HMM and combines GA-RBF neural network to significantly reduce computational complexity while ensuring accuracy, thereby achieving fast and accurate piano playing music note recognition.