Audio to MIDI Conversion
Download
Report
Transcript Audio to MIDI Conversion
Audio Signal MIDI Transcription
DEVON BRYANT
CS 525 SEMESTER PROJECT
Music Transcription
Extraction of “onset”, “duration”, and “pitch”
information from digital audio signals
Two-phased approach
Extract Temporal Information (Onset & Duration)
Extract Frequency Information (Pitch)
Many applications
MIDI representation for low bandwidth
Sheet music score generation
Comparison against DB for copyright or search
Extraction of Temporal Events
Extraction of Temporal Events
Spectral Flux – change in magnitude spectrum
between consecutive frames
2
2
1 N 1
n 1 N 1
n
A(k) x(n)cosk x(n)sin k
N N n 0
N
N n 0
Magnitude Spectrum
Audio File
Frames
FFT
Mag
“Onsets” = window start, “Offsets” = window stop
Extraction of Pitch Information
Extraction of Pitch Information
Process event frames through Fast Fourier
Transform (FFT) to bin frequencies
Use a-priori instrument knowledge to find
fundamental f0 frequencies in spectrum
f (t) a0 sin( f 0t) a1 sin( 2 f 0t) a3 sin( 3 f 0t) ...
Event
Window
Frames
FFT
f0
Estimation
MIDI File
Issues Encountered
Noise artifacts or fluctuations can trigger false onsets
Frequency resolution on shorter events/notes is poor
Results
Single Note Scale
Original audio
Transcribed MIDI
Chords
Original audio
Transcribed MIDI
Short Song
Original audio
Transcribed MIDI
Questions?