Research And Implementation Of Humming To MIDI Automatic Music Transcription System

People can express the melody of mind to the listener through humming. A piece of humming melody can be recorded as a music notation by experienced music transcription experts, and the automatic music transcription system is an intelligent program of simulating the intelligent behavior of experts.This thesis has searched into the automatic music transcription system from both the pitch and melody recognition. The main works are as follow:1. Define auditory standard of humming notes. Humming notes is the basic unit of humming melody. We consider a piece of actual humming as a cognizable auditory component serial. Each component can be presented as a stable pitch with relatively steady sound, or a vibrato pitch with greater volatility, or a portamento pitch which changes along a single direction, or a tone with the start of gas flow, or ventilation or pause and so on.2. Structuring the model of singing note segmentation and pitch estimating. When listening a humming melody, the pitch we hear depends on the contents of humming note and its context. We implement an auto segment program by constructing the augmented transition network. Besides, by observing the humming note pitch confirmed by hearing and the actual note pitch, we creat a model for humming note pitch estimating.3. Build a tonal analysis model based on humming pitch interval. We use a tonal analysis method based on the interval between adjacent humming notes. First, we estimate interval between adjacent humming notes. Then, we estimate the humming melody based on interval series. Humming melody tonal analysis content involves tonal step, temporary changes in the sound, pitch error and transfer analysis.Based on the above research, we implement the Automatic Singing Transcription System——Singing Transcription. The system has good research value and practical value.