Real Time Music Beat Recognition Algorithm Combining Music Style Information and Caching Mechanism

Currently, deep learning algorithms have achieved significant results in image processing and speech recognition, and their application in music beat recognition has gradually become a research focus. Music beat recognition is an important task in music information retrieval, and research is dedicated to achieving efficient real-time music beat recognition. In real-time music beat recognition, convolutional neural networks are used to process spectrograms to obtain music style information, and bidirectional long short-term memory networks and hidden Markov models are used for real-time music beat recognition. At the same time, a caching mechanism is adopted to obtain historical information and achieve global beat recognition. The results showed that the real-time music beat recognition algorithm proposed by the research institute achieved the best recognition performance, with Precision, Recall, F1 value, and Cem_acc value reaching 74.68%, 87.05%, 83.92%, and 83.14%, respectively. Meanwhile, compared with the original model, the real-time music beat recognition model that combines style information recognition and caching mechanism can effectively infer hidden state sequences, and the predicted beat values can correspond to the actual music beat speed, with better beat recognition performance. This method combines the powerful capabilities of deep learning with the advantages of traditional methods, and has achieved significant results in real-time music beat recognition, providing an effective technical foundation for music information retrieval and real-time music analysis.