Research On The Evaluation Methods Of The Acoustical Quality With Music Styles Of Piano Based On Dynamic Fuzzy Neural Network

With the development of society,economy,science and technology,more and more attention has been paid to the study of piano music art combined with information science and computer technology.With the increasing popularity of piano,it has become more and more important to evaluate the piano acoustical quality scientifically and objectively.The objective evaluation methods of piano acoustical quality not only can evaluate the acoustical quality conveniently and effectively,but also has great guiding significance for improving piano manufacturing level.Therefore,this paper studies the related theories and methods of the piano acoustical quality evaluation.Based on the extraction of various features of the piano music signal and the prior knowledge of the subjective evaluation data of the piano acoustical quality,the function of using dynamic fuzzy neural network(D-FNN)for automatic acoustical quality evaluation is realized.The main contents and innovations of this paper are as follows:1.Aiming at the personalized demand scene based on piano music styles,this paper proposes a piano acoustical quality evaluation method combined with music styles and multi-feature extraction.In such a scenario,when evaluating the piano acoustical quality,the method of obtaining the signal of piano sound is changed from obtaining only the sound of some keys or chords in the past to now obtaining an entire musical piece corresponding to some music style;At the same time,the analysis of the features of the piano's sound signals is changed from the analysis of the features of the single keynote to the analysis and recognition of multiple keys and the complex relations between multiple keys.2.According to the relationship between music signal characteristics and piano sound quality,features such as sound length and start-up time are extracted as timedomain features,features such as overtone number and overtone energy are extracted as frequency-domain features,and the matrix eigenvalues of the matrix of the cross-correlation coefficients based on cross-correlation analysis are used as spatial features.According to the characteristics of the keys in different ranges and the differences of role during the performance,with the bass areas,mid-pitch areas and high-pitch areas as a statistical unit,the time-domain features and the frequencydomain features of each key are replaced by the time-domain features and the frequency-domain features in each range.3.Considering the experience knowledge alone can't accurately describe how the features of key signals in a particular style influence the piano acoustical quality,this paper proposes a method for obtaining the logic and rules of reasoning of the complex relationship between music styles,piano acoustical quality,and features by training the dynamic fuzzy neural networks.And using the music signals collected from different pianos and various music styles,Conduct an experiment on the hardware integrated platform independently developed by the laboratory.The experimental results show that the overall performance of this method can achieve the desired effect.