摘要
基于人工智能的乐器识别是音频识别和处理中的重要一环,在近20年来的研究中取得了丰硕的成果,也面临着一些问题.鉴于目前国内的相关研究较少,重点介绍了国际乐器识别领域的研究热点,阐述了应用于乐器识别的特征提取、数据资源、几类常用的数据处理原理及模型.在介绍国际上乐器识别领域的主要成果时,重点阐述了基于机器学习和深度学习的人工智能技术在复调音乐乐器识别和标注中的应用.本综述对于我国开展相应的研究可以提供一些经验和参考.
AI-based musical instrument recognition(MIR)is an important part of audio recognition and processing,and fruitful achievements have been made in recent twenty years,despite of numerous problems.With a relative scarcity of domestic relevant studies,this paper focuses on some hot issues in the international MIR field,and illustrates feature extraction,data resources,as well as data processing principles and models.While introducing major international MIR achievements,this paper emphasizes MIR applications in polyphonic music based on machine learning and deep learning.This literature review can hopefully provide some experience and reference for corresponding researches in China.
作者
谢黛安
XIE Dai-an(Viterbi School of Engineering, University of Southern California, Los Angeles 90007, USA)
出处
《南京工程学院学报(自然科学版)》
2020年第2期66-75,共10页
Journal of Nanjing Institute of Technology(Natural Science Edition)
关键词
乐器识别
音频数据
机器学习
深度学习
musical instrument recognition(MIR)
audio data
machine learning
deep learning