Development of improved automatic music transcription system for the arabian flute (NAY)

The accuracy and reliability of automatic music transcription plays and important role in music information retrieval applications. This paper investigates a previously reported Automatic Music Transcription (AMT) system for the Arabian flute (Nay)with the aim of extending its features and improving its performance. Several critical design factors are considered in the new extension. A hybrid pitch detection method that utilizes both the autocorrelation and Fast Fourier Transform (FFT) is suggested and implemented in the improved AMT system. The signal power is also used to detect octave falls and to perform mapping from intensity to MIDI velocity. Experimental evaluation of the developed system has shown better transcription results when compared to original system.