Research on Vocal Music Identification and Classification Based on Energy Entropy Ratio and AlexNet Model

With the exponential growth of vocal music data in the internet era, effective management of data has become increasingly important. This study focused on vocal music identification and classification. It first reviewed the literature on the development of neural networks in music-related research, along with the challenges of vocal music genre classification. An endpoint detection technique was then proposed, combining the energy entropy ratio method with the pitch estimation filter with amplitude compression algorithm. This approach incorporated a double-threshold mechanism to improve the accuracy of endpoint detection in noisy environments. Additionally, an enhanced model—AlexNet-improved-V1—was developed for recognizing differences in vocal music. Experimental results demonstrated that the proposed methods offered clear advantages in terms of model accuracy and processing efficiency. This research contributes to the integration of vocal music studies and computer technology, holding significant implications for the advancement of intelligent vocal music systems.