Tone Recognition Database of Electronic Pipe Organ Based on Artificial Intelligence

Abstract

In the past few decades, artificial intelligence technology has experienced rapid development, and its application in modern industrial systems has grown rapidly. This research mainly discusses the construction of a database of electronic pipe organ tone recognition based on artificial intelligence. The timbre synthesis module realizes the timbre synthesis of the electronic pipe organ according to the current timbre parameters. The audio time domain information (that is, the audio data obtained by file analysis) is framed and windowed, and fast Fourier transform (FFT) is performed on each frame to obtain the frequency domain information of each frame. The harmonic peak method based on improved confidence is used to identify the pitch, obtain the fundamental tone of the tone, and calculate its multiplier. Based on the timbre parameters obtained in the timbre parameter editing interface, calculate the frequency domain information of the synthesized timbre of each frame, and then perform the inverse Fourier transform to obtain the time domain waveform of each frame; connect the time domain waveforms of different frames by the cross-average method to obtain the time-domain waveform of the synthesized tone (that is, the audio data of the synthesized tone). After collecting the sound of the electronic pipe organ, the audio needs to be denoised, and the imported audio file needs to be parsed to obtain the audio data information. Then, the audio data are frequency-converted and the timbre characteristic information is analyzed; the timbre parameters are obtained through the human-computer interaction interface based on artificial intelligence, and the timbre of the electronic pipe organ is generated. If the timbre effect is not satisfactory, you can re-edit the timbre parameters through the human-computer interaction interface to generate timbre. During the experiment, the overall recognition rate of 3762 notes and 286 beats was 88.6%. The model designed in this study can flexibly generate electronic pipe organ sound libraries of different qualities to meet the requirements of sound authenticity.