基于木材振动特性的月琴声学品质广义回归神经网络预测模型

Research on GRNN Prediction Model for Acoustic Quality of Yueqin Based on Wood Vibration Characteristics

泡桐木始终是制造乐器谐振元件的重要材料,对乐器的音质有着重要的影响。采用广义回归神经网络(General Regression Neural Network,GRNN)建立基于共鸣板振动性能的月琴音质评价模型。以制造出的9把月琴为研究对象,根据月琴的音质评价以及制备月琴的共鸣板信息,提出月琴音质的预测模型。在180组数据中,随机抽取135组数据进行训练,其余45组数据进行验证。使用主成分分析方法、GRNN建立月琴声学质量评价模型,并进行仿真预测。结果表明,基于共鸣板的振动特性,利用Matlab仿真可以实现对月琴音质的预测,预测的准确率可达到91.41%。此外,研究还表明,泡桐木共鸣板的动态弹性模量、声辐射阻尼系数、弹性模量、剪切模量比、声阻抗,损耗角正切和声转化率等参数均是影响其制备成品月琴声学质量的重要因素。

Paulownia has usually been an important material for making resonant components of musical instruments,which has a significant influence on the sound quality of musical instruments.This study utilized a generalized regression neural network(GRNN)to develop the sound quality evaluation model of Yueqin based on the vibration performance of the soundboard.In this study,nine Yueqins were fabricated,and a prediction model for the sound quality of Yueqins was proposed based on their sound quality evaluation and the soundboard information of prepared Yueqins.Out of a total of 180 sets of data,135 sets of data were randomly selected for training and the remaining 45 sets of data were used for validation.A model for evaluating the acoustic quality of Yueqin instruments was established using principal component analysis method and GRNN,and simulation prediction was performed.The results showed that based on the vibration characteristics of the soundboard,the prediction of the Yueqin sound quality can be achieved by using the Matlab simulation,and the accuracy of the prediction can reach 91.41%.In addition,this study demonstrated that the dynamic elastic modulus,acoustic radiation damping coefficient,elastic modulus,elastic and shear modulus ratio,acoustic impedance,loss tangent angle,and acoustic conversion efficiency of Paulownia wood resonator plates were all key factors influencing the acoustic quality of the finished Yueqin during its preparation.