Construction of a Quantitative Evaluation Model for Bamboo Flute Timbre Characteristics Based on Spectral Analysis

The evaluation of bamboo flute timbre traditionally relies on subjective auditory experience, making it difficult to establish standardized teaching and objective assessment. To address this limitation, this paper constructs a quantitative evaluation model for bamboo flute timbre characteristics based on spectral analysis. A total of 240 audio samples from Southern and Northern school performers are recorded, and key parameters including static harmonic features, transient noise intensity, and dynamic spectral-centroid drift are extracted using short-time Fourier transform and an adaptive double-threshold algorithm. Principal component analysis is used for dimensionality reduction and feature fusion, while support vector regression is applied to obtain quantitative timbre scores. The results show that the model scores are highly consistent with expert scores, with a correlation coefficient of R = 0.974. The model effectively distinguishes Northern and Southern school styles, with an inter-class/intra-class distance ratio of 3.03, and sensitively identifies dynamic timbre changes (p < 0.001). The study provides an objective quantitative basis for timbre perception, teaching feedback, and digital inheritance of bamboo flute performance.