Research Of Music Emotion Recognition Method Based Music Feature Vector Space

Nowadays, computer technology and digital music technology are evolving rapidly. Asan independent branch of art which combines art and IT, computer music becomes a focus forresearchers. The essential characteristic of the music is its inherent emotion. Music emotionrecognition is a type of automatic identification of musical works based on the mapping ofmusical features and emotional relationship. Automatic emotion recognition of digitizedmusic is important for enhancing the emotion of the man-machine interactive capabilities ofthe computer.In this paper, we choose MIDI music files as a sample of the study. We make acomprehensive and detailed description of the characteristics of the music, by following theclassical music theory and research of music psychology, cognitive psychology and musicaesthetics. We establish a system of music features for computer understanding andexpression. In the process of feature extraction for MIDI music, we make a mathematicalexpression for rhythm, tempo, melody, and other high-level features. This lay the foundationfor the expert knowledge of MIDI music.This paper presents a musical characteristics and musical emotion mapping modelapplied to the emotion analysis of digital music. We proposed the workflow and architectureof music emotion recognition system, and achieve the computer music emotion recognitionand expression of music emotion by using expert system and BP neural network. This paperproposes a rule-based music emotion cognition expert system which uses production rules anduncertain reasoning techniques to analyze the musical emotion. We also design the input layerand output layer of BP neural network for music emotion recognition according to musicalfeatures and emotional mapping model. We get effective training samples and simulationexperiments in MATLAB through the experts selection of60samples, and basically achievedthe desired results.