Audio classification using attention-augmented convolutional neural network

Abstract

2018) Audio classification using attention-augmented convolutional neural network. Published by: Elsevier URL: http://dx. A B S T R A C TAudio classification, as a set of important and challenging tasks, groups speech signals according to speakers' identities, accents, and emotional states. Due to the high dimensionality of the audio data, task-specific handcrafted features extraction is always required and regarded cumbersome for various audio classification tasks. More importantly, the inherent relationship among features has not been fully exploited. In this paper, the original speech signal is first represented as spectrogram and later be split along the frequency domain to form frequency-distributed spectrogram. This paper proposes a task-independent model, called FreqCNN, to automaticly extract distinctive features from each frequency band by using convolutional kernels. Further more, an attention mechanism is introduced to systematically enhance the features from certain frequency bands. The proposed FreqCNN is evaluated on three publicly available speech databases thorough three independent classification tasks. The obtained results demonstrate superior performance over the state-of-the-art.

Audio classification using grasshopper‐ride optimization algorithm‐based support vector machine

“…Moreover, the classification accuracy of the method was not good in case of the presence of environmental noises but was effective in the case of clean environments. Wu et al [1] developed an attention-augmented CNN that enhanced the features generated from the frequency bands. The drawback of the method was regarding the higher number of the local frequency segments as this would increase the parameters in the model, affecting the performance of the system.…”

“…Audio classification differentiates the audio using the emotion, identity, accents, and other parameters of the speakers, which when performed effectively, can contribute the tasks, like translation from speech-to-speech and automatic recognition of the speech [1]. For example, the emotions of the sound are employed for the translation of the utterances spoken in a language to another [2] that facilitates the ability to handle the non-linguistic data and ensures the practical way of translation.…”

“…Most of the audio classification methods utilize the lowdimensional features for classification since the original audio suffers from a high dimensionality problem [11] [12] [13]. For the extraction of the signal details, including energy, time, and frequency from the original audio data, spectrograms are employed [1] [14] [15]. To minimize the input space dimensions, the feature extraction is progressed using the linear prediction cepstrum coefficient and Mel-frequency cepstral coefficients (MFCCs) [1].…”

“…For the extraction of the signal details, including energy, time, and frequency from the original audio data, spectrograms are employed [1] [14] [15]. To minimize the input space dimensions, the feature extraction is progressed using the linear prediction cepstrum coefficient and Mel-frequency cepstral coefficients (MFCCs) [1]. Though they possess complex sound proportions, many details are common in the audio scenes belonging to the same category.…”

Weakly Labelled Audio Tagging Via Convolutional Networks with Spatial and Channel-Wise Attention

“…In addition, batch normalization [17] is attached to the convolutional layer for accelerating the learning. Secondly, since the event occurring timestamp is important to distinguish between events, we also consider local attention which aims to capture the onset and offset of the events by taking the local patch features as input [18,19]. Specifically, along the frequency axis with the interval set to 1, the global spatial descriptor S is split into local parts, denoted as local descriptors s 1 , s 2 , .…”