Abstract
Speaker Identification using i-vector has gradually been replaced by speaker Identification using deep learning. Speaker Identification based on Convolutional Neural Networks (CNNs) has been widely used in recent years, which learn low-level speech representations from raw waveforms. On this basis, a CNN architecture called SincNet proposes a kind of unique convolutional layer, which has achieved band-pass filters. Compared with standard CNNs, SincNet learns the low and high cutoff frequencies of each filter.This paper proposes an improved CNNs architecture called LineNet, which encourages the first convolutional layer to implement more specific filters than SincNet. LineNet parameterizes the frequency domain shape and can realize band-pass filters by learning some deformation points in frequency domain. Compared with standard CNN, LineNet can learn the characteristics of each filter. Compared with SincNet, LineNet can learn more characteristic parameters, instead of only low and high cutoff frequencies. This provides a personalized filter bank for different tasks. As a result, our experiments show that the LineNet converges faster than standard CNN and performs better than SincNet.
Abstract (translated)
URL
https://arxiv.org/abs/2105.14826
