Machine learning via multimodal signal processing

This paper proposes a methodology for recognition of vocal music (Byzantine music) via multi-modal signals processing. A sequence of multi-modal signals is captured from the expert's (teacher) and student's hymns performances, respectively. The machine learning system is trained using the values of particular features which are extracted from the captured multi-modal signals. After the system is being trained then it becomes able to recognize any hymn performance from the corpus. Training and recognition takes place in real time by utilizing machine learning techniques. The evaluation of the system was carried out with the cross - validation statistical method Jackknife, giving promising results.