Real-time classification via sparse representation in acoustic sensor networks

Abstract

Acoustic Sensor Networks (ASNs) have a wide range of applications in natural and urban environment monitoring, as well as indoor activity monitoring. In-network classification is critically important in ASNs because wireless transmission costs several orders of magnitude more energy than computation. The main challenges of in-network classification in ASNs include e↵ective feature selection, intensive computation requirement and high noise levels. To address these challenges, we propose a sparse representation based featureless, low computational cost, and noise resilient framework for in-network classification in ASNs. The key component of Sparse Approximation based Classification (SAC),`1 minimization, is a convex optimization problem, and is known to be computationally expensive. Furthermore, SAC algorithms assumes that the test samples are a linear combination of a few training samples in the training sets. For acoustic applications, this results in a very large training dictionary, making the computation infeasible to be performed on resource constrained ASN platforms. Therefore, we propose several techniques to reduce the size of the problem, so as to fit SAC for in-network classification in ASNs. Our extensive evaluation using two real-life datasets (consisting of calls from 14 frog species and 20 cricket species respectively) shows that the proposed SAC framework outperforms conventional approaches such as Support Vector Machines (SVMs) and k-Nearest Neighbor (kNN) in terms of classification accuracy and robustness. Moreover, our SAC approach can deal with multi-label classification which is common in ASNs. Finally, we explore the system design spaces and demonstrate the real-time feasibility of the proposed framework by the implementation and evaluation of an acoustic classification application on an embedded ASN testbed.

Novel activity classification and occupancy estimation methods for intelligent HVAC (heating, ventilation and air conditioning) systems

“…the fundamental principles of compressive sensing and allows for further energy savings by classifying directly on the compressed signal without needing to reconstruct the original signal. It has been widely used for multifarious classifications including face, facial expression, gait, physical activity, abnormal scene and many more [34,7,37,39,36].…”

Face recognition on smartphones via optimised Sparse Representation Classification

“…They have been applied to speed up background subtraction on embedded system [34] and cross-correlation computation in sensor networks [35]. In [5], SRC is used for acoustic classification and a column reduction procedure is proposed to reduce the dimension of 1 minimisation. Note that column reduction in [5] is complementary to the techniques of projection matrix optimisation and compressed resi-duals proposed in this paper; all three can be applied to improve the performance of SRC.…”

“…In [5], SRC is used for acoustic classification and a column reduction procedure is proposed to reduce the dimension of 1 minimisation. Note that column reduction in [5] is complementary to the techniques of projection matrix optimisation and compressed resi-duals proposed in this paper; all three can be applied to improve the performance of SRC. Other application of random projection matrix is to enable efficient moisture data collection in sensor networks [36] and privacy preservation of voice data [30].…”

Sensor-based smart recognition system For wearable devices

“…(NB). The intuition for using SRC is that it has shown better performance than traditional classification methods (e.g., SVM and KNN) in recognition tasks such as face recognition [124,140] and voice recognition [137]. SRC is known to be robust to noise because of its use of 1 optimization [124].…”

“…Moreover, SRC has also been applied in sensor areas to solve a range of recognition tasks because it is known to be robust to noise. For example, Wei et al [137] developed an acoustic classification system on wireless sensor networks by applying SRC to improve the recognition accuracy. Shen et al [124] proposed opti-SRC by optimizing the random matrix used in SRC to increase the performance of face recognition system in smartphones.…”