This book constitutes the refereed proceedings of the 12th Conference, CSMT 2025, held in Hohhot, China, on November 79, 2025.

The 17 full papers presented in this volume were carefully selected and reviewed from 68 submissions. They are classified into the following topical sections: Audio Processing and Music Generation Technologies; Acoustics and Sound Source Localization Technologies; and Research on the Integration of Music Technology and Culture.

Inhalt

.- Audio Processing and Music Generation Technologies

.- MVM-ASR: A Lightweight Model for Music Efficient Audio Super-Resolution.
.- A Survey: Neural Vocoders and Their Impact on Music Synthesis.
.- CLAP-Reward Guided Diffusion Fine-Tuning for an Audio Generation Model.
.- Audio Super Resolution with Three-dimensional Modeling in Spectral Domain.
.- Heart Sound Classification via Adaptive Feature Selection and Two-Level Clustering Based on the Ising Model.

.- Acoustics and Sound Source Localization Technologies

.- Simultaneous DOA Estimation and Multi-Channel Speech Enhancement based on Joint Learning Network.
.- Source DOA Estimation Based on Polynomial Eigenvalue Decomposition and First Order Relative Harmonic Coefficients.
.- Active Noise Control Based on VBAP and Grey Wolf Optimizer.
.- Modified Steering Vector: An Analytical Framework for 3D Microphone Array Evaluation and Design.
.- HRTF of the Auditory Perceptual Average Ear: Analysis and Evaluation.
.- Sound Event Detection for Driving Safety Based on Time-Frequency Feature Fusion.
.- Gaussian-Weighted Ray Tracing for Efficient Microphone Signal Modeling in Reverberant Environments.

.- Research on the Integration of Music Technology and Culture

.- SaOMR: End-to-End Sparse Attention-based Full-Page Optical Music Recognition.
.- Oral Tradition-Encoded NanyinHGNN: Integrating Nanyin Music Preservation and Generation through a Pipa-Centric Dataset.
.- A Fast and Contactless Approach to Resonance Feature Acquisition of the Guqin.
.- Structural Complexity Prediction of Chinese Traditional Guzheng Variations Based on Deep Learning.