Deep learning has become a predominant method for solving data analysis problems in virtually all fields of science and engineering. The increasing complexity and the large volume of data collected by diverse sensor systems have spurred the development of deep learning methods and have fundamentally transformed the way the data are acquired, processed, analyzed, and interpreted. With the rapid development of deep learning technology and its ever-increasing range of successful applications across diverse sensor systems, there is an urgent need to provide a comprehensive investigation of deep learning in this domain from a holistic view. This survey paper aims to contribute to this by systematically investigating deep learning models/methods and their applications across diverse sensor systems. It also provides a comprehensive summary of deep learning implementation tips and links to tutorials, open-source codes, and pretrained models, which can serve as an excellent self-contained reference for deep learning practitioners and those seeking to innovate deep learning in this space. In addition, this paper provides insights into research topics in diverse sensor systems where deep learning has not yet been well-developed, and highlights challenges and future opportunities. This survey serves as a catalyst to accelerate the application and transformation of deep learning in diverse sensor systems.

/pdf/deep-learning-in-diverse-intelligent-sensor-based-systems-1tdo1odr.pdf

Deep Learning in Diverse Intelligent Sensor Based Systems

Log-Mel spectrogram for the convolutional neural network (CNN) and wavelet time scattering for Ensemble of subspace discriminant classifiers is used for classifying acoustic scenes with human speech. The Tampere University of Technology (TUT) Acoustic Scenes dataset is used to demonstrate the feasibility of the proposed model. Comparisons are performed with the baseline model in the TUT 2017 dataset used for Detection and Classification of Acoustic Scenes and Events (DCASE) 2017 Challenge-Task 1. The fused model shows good acoustic classification accuracy of 79.43%. The proposed late fusion of multi-model using CNN and ensemble classifiers exhibits 18.4% higher accuracy than the baseline model with just CNN.

Late Fusion of Convolutional Neural Network with Wavelet-Based Ensemble Classifier for Acoustic Scene Classification

Computer audition (i.e., intelligent audio) has made great strides in recent years; however, it is still far from achieving holistic hearing abilities, which more appropriately mimic human-like und...

https://journals.sagepub.com/doi/pdf/10.1177/23312165211046135

New Avenues in Audio Intelligence: Towards Holistic Real-life Audio Understanding.

MobileNet is widely used for Acoustic Scene Classiﬁcation (ASC) in embedded systems. Existing works reduce the complexity of ASC algorithms by pruning some components, e.g. pruning channels in the convolutional layer. In practice, the maximum proportion of channels being pruned, which is deﬁned as Ratio of Prunable Channels ( R PC ), is often decided empirically. This paper proposes a method that determines the R PC by simple linear regression models related to the Sparsity of Channels ( S C ) in the convolutional layers. In the experiment, R PC is examined by removing inactive channels until reaching a knee point of performance decrease. Simple methods for calculating the S C of trained models and resulted R PC are proposed. The experiment results demonstrate that 1) the decision of R PC is linearly dependent on S C and the hyper-parameters have a little impact on the relationship; 2) MobileNet shows a high sensitivity and stability on proposed method.

Determining Ratio of Prunable Channels in MobileNet by Sparsity for Acoustic Scene Classification

Acoustic Scene Classification (ASC) algorithms are usually expected to be deployed in resource-constrained systems. Existing works reduce the complexity of ASC algorithms by pruning some components, e.g. pruning channels in neural network. In practice, neural networks are often trained with sparsification such that unimportant channels can be found and further pruned. However, little efforts have been made to explore the the impact of channel sparsity on neural network pruning. To fully utilize the benefits of pruning for ASC, and to make sure the model performs consistently, we need a more profound comprehension of channel sparsification and its effects. This paper examines the internal weights acquired by convolutional neural networks that will undergone pruning. The study discusses how these weights can be utilized to create a novel metric, Weight Skewness (WS), for quantifying the sparsity of channels. We also provide a new approach to compare the performance of different pruning methods, which balances the trade-off between accuracy and complexity. The experiment results demonstrate that 1) applying higher channel sparsity to models can achieve greater compression rates while maintaining acceptable levels of accuracy; 2) the selection of pruning method has little influence on result 1); 3) MobileNets exhibit more significant benefits from channel sparsification than VGGNets and ResNets.

Exploring the Effects of Channel Sparsity on Neural Network Pruning for Acoustic Scene Classification

Different audio environments require different settings in hearing aid to acquire high-quality speech. Manual tuning of hearing aid settings can be irritating. Thus, hearing aids can be provided with options and settings that can be tuned based on the audio environment. In this paper we provide a simple sound classification system that could be used to automatically switch between various hearing aid algorithms based on the auditory related scene. Features like MFCC, Mel-spectrogram, Chroma, Spectral contrast and Tonnetz are extracted from several hours of audio from five classes like “music,” “noise,” “speech with noise,” “silence,” and “clean speech” for training and testing the network. Using these features audio is processed by the convolution neural network. We show that our system accomplishes high precision with just three to five second duration per scene. The algorithm is efficient and consumes less memory footprint. It is possible to implement the system in digital hearing aid.

S Vidhya

Papers

Acoustic Scene Classification in Hearing aid using Deep Learning