Pitch estimation for musical note recognition using Artificial Neural Networks
Jose de Jesus Guerrero-Turrubiates, Sheila Esmeralda Gonzalez-Reyna, Sergio Ledesma, Juan Gabriel Avina‐Cervantes · 2014
Pitch estimation has increased its importance due to the wide variety of applications in different fields, e.g. speech and voice recognition, music transcription, to name a few. Musical signals may contain noise and distortion, therefore pitch detection results can be erroneous. In this paper, a musical note recognition system based on harmonic modification and Artificial Neural Network (ANN) is proposed. At first, downsampling is applied to convert the signal from 44,100 Hz sampling rate to 2,100 Hz. Fast Fourier Transform (FFT) is used to obtain the signal spectrum; Harmonic Product Spectrum (HPS) algorithm is implemented to enhance the fundamental frequency amplitude. Then a dimensionality reduction method based on variances, is used to extract relevant information from the input signal. In the present work, audio signals were taken from a proprietary database that was constructed using an electric guitar as audio source. The classification is performed by a feed-forward neural network or Multi-Layer Perceptron (MLP). Experimental results present accurate classification with few processing of the input signal. Besides the proposed approach presents enough robustness to classify musical notes coming from different musical instruments.