Chord Recognition Method Based on Robust Scale Contour Features and Vector Machine

Abstract

The invention relates to automatic chord recognition, in particular to a chord recognition method based on robust musical scale contour features and a vector machine, wherein the chord recognition method is capable of fast and accurately recognizing chord by extracting robust musical scale contour chord characteristics. The chord recognition method based on robust musical scale contour features and the vector machine includes the following steps that 1) windowing preprocessing is carried out on original audio signals; 2) discrete cosine transform is carried out on a framing result to obtain a standard audio spectrum matrix S of the original signals; 3) a globally optimal solution is obtained through a convex optimization problem; 4) matrix mapping is carried out to obtain robustness PCP characteristics; 5) a measure learning method is adopted to optimize a gaussian kernel function supporting the vector machine; 6) training data is used for training the measure learning support vector machine, and parameters in the measure learning support vector machine are determined; 7) the trained measure learning support vector machine is used for recognizing test data to obtain the final recognition rate. The chord recognition method is mainly used on the automatic chord recognition occasion.

Description

technical field [0001] The present invention relates to automatic recognition of chords, in particular to a chord recognition method based on Robust PitchClass Profiles (RPCP) and metric learning Support Vector Machine (mlSVM). Background technique [0002] Chord recognition is one of the important research problems in music signal processing. It plays an important role in the fields of song cover recognition, audio matching and music recommendation system. As an important part of music, chords are composed of more than three tones superimposed according to the relationship of thirds, which fully expresses the content and characteristics of a piece of music, and plays an important role in the cognition of music. Therefore, the structural features of chords in the frequency domain and chord recognition is a key issue in computer music signal processing. [0003] It is generally believed that chord recognition is one of the central tasks of music information retrieval, and it...

Images