Model for Polyphonic Piano Transcription
This paper presents a discriminative model for polyphonic piano transcription. Using support vector machines to classify note instances, it is able to accurately transcribe both synthesized and real piano recordings. Its accuracy achieved 68% on a newly generated test set.
Convolutional sparse coding approximates the music waveform. Piano note waveforms are pre-recorded for each piano, fixed during transcription, and estimated during post-processing. This method estimates pitches simultaneously and models the temporal evolution of the notes.
The proposed model is trained on audio recorded from the Yamaha Disklavier piano. The audio is sampled at 100 Hz. The test audio has 54 test files, corresponding to the notes A0-C8 on a piano. The test files contain 648, 000 frames in total at the test time.
A Discriminative Model for Polyphonic Piano Transcription
This algorithm begins by analyzing note onset. Two notes are considered to belong to the same onset if they are played within 32ms. Then, the recording is transformed by performing a Hanning windowing followed by a Constant Q Transform.
The algorithm also uses pitch salience estimates to map the notes to chords. The algorithm is very efficient and can even be used to process streaming audio. It has been successfully tested on the first eight Beatles albums. It also has a wide range of applications. It can be used to analyze musical recordings, retrieve music information from large music databases, and create interactive music systems.
Subscribe to Tartalover Youtube Channel for more Piano Transcriptions videos
This method is effective for detecting multiple pitches in multi-pitch piano recordings. Using note event metrics, it can also detect note onsets with high accuracy. Generally, note event metrics are assumed to be accurate if they fall within +-50 ms of ground truth onset. However, note event metrics are difficult to implement in real-time and require a large amount of time.