基于粒子群算法优化极限学习机的钢琴曲类型识别  

Piano Music Type Recognition based on extreme learning machine improved by particle swarm optimization algorithm

作　　者：王力博[1] Wang Libo(Xi'an Jiaotong Engineering Institute,Xi'an,Shaanxi 710300)

机构地区：[1]西安交通工程学院,陕西西安710300

出　　处：《现代科学仪器》2022年第2期198-201,211,共5页Modern Scientific Instruments

基　　金：西安交通工程学院2020年中青年项目(编号:20KY-53)。

摘　　要：为提高钢琴曲类型识别的精度,提出一种基于MFCC和PSO-ELM的钢琴曲类型识别模型。首先,提取不同音乐风格的MFCC系数矩阵作为ELM的输入向量,钢琴曲类型作为ELM的输出向量,建立钢琴曲ELM识别模型;然后,使用PSO对ELM模型进行优化,获得最佳的输入层权重以及隐含层阈值,建立GWO-ELM的钢琴曲识别模型。研究结果表明,与KNN、FA-ELM、GA-ELM和DE-ELM相比较,PSO-ELM模型的钢琴曲类型识别精度最高,为钢琴曲类型识别提供了新的方法。In order to improve the accuracy of piano music type recognition,a piano music type recognition model based on MFCC and PSO-ELM is proposed.Firstly,the MFCC coefficient matrix of different musical styles is extracted as the input vector of ELM,and the piano music type is taken as the output vector of ELM,and the recognition model of ELM is established.Secondly,PSO is used to optimize the weights of the input layer and the threshold of the hidden layer of ELM,a piano music recognition model based on GWO-ELM was established.Compared with KNN,FA-ELM,GA-ELM,GA-ELM and DE-ELM,PSO-ELM model has the highest accuracy in piano music type recognition,which provides a new method for piano music type recognition.

关 键 词：粒子群算法 极限学习机 MEL频率倒谱系数 钢琴曲 音乐风格 

分 类 号：TH124[机械工程—机械设计及理论]