Musical feature matrices were developed with computer technology and have been most fully developed in modern music composition in recent years. This paper extracts and analyzes the timbral feature matrix and tone intensity change envelope curve of Debussy’s piano piece using the spectral analysis method of short-time Fourier transform. Also through the MIDI file in the extraction of each note length, pitch, pitch velocity, channel and other information, the establishment of note feature matrix, for digital music creation to lay the foundation. The experimental results and analysis show that the method in this paper can accurately recognize the time value of the piano music score, and its error values are all within the range of (-5%, 5%). The BER of the proposed algorithm in this paper is lower, the signal-to-noise ratio of the recognition algorithm is smaller, and there is a clear delineation of audio signal features. The tonal color and acoustic characterization of three piano pieces, Prelude, Moonlight, and Minuet, are completed. Finally, the sound color of “Moonlight” is improved with sine wave musical notes, which lays the foundation for modern music creation.
