Research on the Extraction of Digital Features of Vocal Performance Styles and the Construction of a Database for Inheritance and Protection

With the paradigm shift in vocal art research, transforming highly abstract and experience-dependent vocal performance styles into quantifiable and analyzable digital objects has become an important issue in music information processing and digital heritage preservation. This study focuses on digital feature extraction and database construction for vocal performance styles, aiming to build a multi-dimensional feature fusion framework and a searchable database system. The study first establishes a multimodal data acquisition standard covering audio signals, physiological parameters, and body posture movements. Then, short-time Fourier transform, fundamental frequency tracking, surface electromyography analysis, and computer vision technology are used to deconstruct and quantify performance styles from acoustic, physiological, and physical dimensions. Based on these features, a vocal performance style database supporting multidimensional retrieval, similarity matching, and visual analysis is designed and implemented. Experimental results show that vibrato frequency, formant position, harmonic-noise ratio, energy dynamic range, and posture-related features can effectively distinguish different vocal schools and individual performance characteristics. The database achieves an average response time of 1.32 seconds under a concurrent retrieval environment, indicating acceptable retrieval efficiency for teaching and research applications. The proposed framework provides an operable technical path for the digital preservation of vocal art and offers data support for vocal teaching, style comparison, and AI-assisted creation. Its time-frequency feature extraction and waveform-indexing mechanism also provide methodological reference for other wave-based signal archives, including acoustic and electromagnetic signal data management.