CGCNN: COMPLEX GABOR CONVOLUTIONAL NEURAL NETWORK ON RAW SPEECH - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2020

CGCNN: COMPLEX GABOR CONVOLUTIONAL NEURAL NETWORK ON RAW SPEECH

Mohamed Morchid

Abstract

Convolutional Neural Networks (CNN) have been used in Automatic Speech Recognition (ASR) to learn representations directly from the raw signal instead of hand-crafted acoustic features, providing a richer and lossless input signal. Recent researches propose to inject prior acoustic knowledge to the first convolutional layer by integrating the shape of the impulse responses in order to increase both the interpretability of the learnt acoustic model, and its performances. We propose to combine the complex Gabor filter with complex-valued deep neural networks to replace usual CNN weights kernels, to fully take advantage of its optimal time-frequency resolution and of the complex domain. The conducted experiments on the TIMIT phoneme recognition task shows that the proposed approach reaches top-of-the-line performances while remaining interpretable.
Fichier principal
Vignette du fichier
gabor_complex_cnn_final.pdf (317.03 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02474746 , version 1 (11-02-2020)

Identifiers

  • HAL Id : hal-02474746 , version 1

Cite

Paul-Gauthier Noé, Titouan Parcollet, Mohamed Morchid. CGCNN: COMPLEX GABOR CONVOLUTIONAL NEURAL NETWORK ON RAW SPEECH. ICASSP 2020, May 2020, Barcelona, Spain. ⟨hal-02474746⟩

Collections

UNIV-AVIGNON LIA
247 View
241 Download

Share

Gmail Facebook X LinkedIn More