Design and Development of Real‐Time Music Transcription Using Digital Signal Processing

This chapter focuses on designing a real‐time monophonic music transcriber using digital signal processing (DSP). Music information retrieval is one of the challenging tasks in music transcription. Short‐time Fourier transform (STFT) and continuous wavelet transform (CWT) are used on WAV files for chroma extraction, image processing (IP) for note extraction, and musical instrument digital interface (MIDI) conversion for audio‐feature extraction (e.g. a beat, pitch, time duration), resulting in sequential keys of the piano (in chromatic form) as the output. This mechanism's capability is verified on a database: The NSynth dataset contains 305 979 music files with various musical genres and instruments. Also, this dataset possesses different pitch classes and frequency ranges. The proposed technique in this chapter works for all keys from C4 to A7, covering four octaves.