On the Design of Solfeggio Audio Machine Assessment System

Abstract

This paper presents an approach of solfeggio audio machine assessment based on multiple audio spectrum features. In this machine assessment system, we analyze 13 spectrum audio features of solfeggio audios, and build a machine assessment model based on the scoring results of music experts, aiming to score the solfeggio audio system at the professional level based on machine learning techniques. An effective way is designed to implement the system with high generalization ability in this paper, with deploying 150 practice audio files marked with expert scores for training and testing the model. Various features are extracted to train predictors, which access the distances between practice audio files and their target audio files. In this way, these predictors can mark unknown audio files reliably. The training and testing process are described in detail, and the test results testify that our system can provide expert level scoring.