Applications of Stochastic Machine Learning and Statistical Signal Processing Approaches to Automatic Music Transcription and Visualisation

将复杂的复调和多音色演奏转录为乐谱的过程是一项众所周知的艰巨任务，需要能够区分许多乐器线条，这些线条通常只能通过音色，频率和波形特征的显着细微变化来分离。拟议的研究涉及调查几种用于转录过程自动化的新技术，结合对通过深度学习技术获得的自动可视化实用程序的探索，以及对以前获得的方法的局限性的分析。

The process of transcribing complex polyphonic and polytimbral performances to musical notation is a notoriously arduous task, requiring the ability to distinguish numerous instrumental lines, separable often only via remarkably subtle variances in timbre, frequency, and waveform characteristics. The proposed research involves investigating several novel techniques for the automation of the transcription process, combined with the exploration of automated visualisation utilities derived through deep learning techniques, and analysis of the limitations of previously derived methods.