Research On Music Emotion Recognition Method Based On Machine Learning

Music plays an important role in human history,especially in the digital age.Today's music is growing exponentially,and at the same time,the need to organize classifications and retrieve music is increasing.Musical emotion based classification and retrieval is different from traditional music text based classification and retrieval,paying more attention to the emotional expression of the creator and the unique characteristics of music in psychology.It is an indispensable part of the personalized needs of the user population,and so it is getting more and more attention.Music emotion recognition is the classification and recognition of music by emotion.This paper analyzes the current research status of music emotion recognition at home and abroad,and summarizes the emotion model,data set,music features,machine learning algorithm and system framework in the existing music emotion recognition research.According to these characteristics,the machine learning is chosen to perform musical emotion recognition.With the rapid development of artificial intelligence,research on machine learning is also increasing.The process of machine learning is to train the model through existing data and then predict the results with t he trained model.Machine learning allows the computer to simulate the process of learning,and then predicts and judges the unknown through empirical knowledge.This paper chooses to use the classic k-Nearest Neighbor(KNN)regression and Support Vector Machine(SVM)algorithm to realize the realization of music emotion recognition of dimensional and discrete models.Due to the higher uncertainty of the dimensional model,the KNN regression algorithm is not effective,and there is still much room for improvement.Other auxiliary algorithms or other music information can be used to improve the recognition accuracy.Meanwhile,the influence of the neighbors on the recognition accuracy is compared in the experiment.If the number of neighbors is too large or too small,the recognition performance will be worse,so the appropriate neighbor number must be selected.In the discrete model,the SVM algorithm performs well.In the experiment,the recognition effect under different kernel functions is compared.It is found that the choice of kernel function has a great influence on the recognition result.To obtain a higher recognition rate,we must choose the appropriate one.Kernel function.Finally,this paper summarizes the current problems in music emotion recognition through machine learning,and forecasts the future of music emotion recognition.