An investigative team at the University of Van in Spain, Spain, has defined an automated system that determines the spectral pattern of a musical note. The model is used to create excited vocabulary, which is combined with the algorithm for the model. The system determines which note and which information to convert into identifiable form. Looking at the WAV recording file, the software can generate MIDI transcript.
Automated music transcription can help music experts analyze sound patterns, music downloads, and various audio sources, said Julio Jose Jose Carabis, co-author and researcher at the University of Engineering's Department of Telecommunications Engineering.
Details of the procedure were published in _EE transactions for audio, speech and language processing. This system is customizable, which means you can play any device from dulcimers to doggardido. From now on, it only works on one device at a time, but researchers think that this method could be expanded to include multiple devices running simultaneously. Other transcription music devices use a database and are trained to recognize specific notes, just as a spectrometer is trained to identify the spectrum of certain chemical compounds. But the Spanish device learns by itself, and creates its own dictionary.
The sound spectrum represents the sound in terms of the amount of vibration at each individual frequency. The note's harmonic energy distribution defines its color patterns. Using this information, the system creates vocabularies for sound. Identify notes even when device type, musician, type of music, or studio conditions are different.
“Another benefit of this method is that it does not require prior training with the music database,” said Krabis.
