Robust multipitch estimation for the analysis and manipulation of polyphonic musical signals

Audio demonstrations page


This page contains the audio demonstrations that accompany the conference paper

Klapuri, Virtanen, Holm. (2000). " Robust multipitch estimation for the analysis and manipulation of polyphonic musical signals, " In Proc. COST-G6 Conference on Digital Audio Effects, December 7-9, 2000, Verona, Italy.

Transcription example

SIGNAL 1 original:
Original acoustic signal that was fed to a transcriber. A Finnish folksong "Taivas on sininen ja valkoinen".

SIGNAL 1 piano:
The original acoustical signal has been automatically transcribed, and resynthesized using piano sounds. Some errors have occurred, but the piece can still be easily recognized.

SIGNAL 1 violin:
Same as above, but violin voices have been used in resynthesis.

More transcription examples (somewhat older stuff)

All the transcriptions below have been done from wave to MIDI, with no knowledge of the polyphony or instruments in the performance. All notes from 30Hz to 4300Hz are assumed equally probable to appear, and musical context is not utilized.

Original Transcribed, then resynthesized Transcribed, modified, resynthesized
Yesterday.wav Yesterday transcribed.wav  
Bach's Inventio 8 for two voices.wav Inventio 8 transcribed.wav Inventio 8 modified.wav

Sound separation and timbre parameter extraction

SIGNAL 2 original:
A sequence of musical chords, varying the instrument.

SIGNAL 2 arpeggio:
The acoustical signal has been automatically transcribed, the sounds have been separated from the mixture, and the same sequence is replayed in arpeggion, i.e., breaking the chords to successive notes. Limitations of the sound separation stage appear as distortions in the separated timbres.

SIGNAL 2 sinusoidal arpeggio:
Same as above, but now the separated sounds are only used to drive the amplitude parameter of a sine wave, the frequency of which has been set according to the detected pitches in the original signal.

SIGNAL 2 sinusoidal chords:
The original sequence of chord played with sinusoidal timbres, the amplitude envelopes of which are driven by the parameters that were extracted from the original.

Applying effects to a one sound in a mixture

Original acoustic mixture (G7 major chord), followed by a version where tremolo effect has been applied to one of the sounds, first to the note D, and then to the note F.


Audio Research Group

Last modified: Tue Dec 11 14:03:40 EET 2001 - Anssi Klapuri, klap @ cs tut fi