Schedule-Based Cooperative Multi-agent Reinforcement Learning for Multi-channel Communication in Wireless Sensor Networks

Wireless sensor networks (WSNs) have become an important component in the Internet of things (IoT) field. In WSNs, multi-channel protocols have been developed to overcome some limitations related to the throughput and delivery rate which have become necessary for many IoT applications that require sufficient bandwidth to transmit a large amount of data. However, the requirement of frequent negotiation for channel assignment in distributed multi-channel protocols incurs an extra-large communication overhead which results in a reduction of the network lifetime. To deal with this requirement in an energy-efficient way is a challenging task. Hence, the Reinforcement Learning (RL) approach for channel assignment is used to overcome this problem. Nevertheless, the use of the RL approach requires a number of iterations to obtain the best solution which in turn creates a communication overhead and time-wasting. In this paper, a Self-schedule based Cooperative multi-agent Reinforcement Learning for Channel Assignment (SCRL CA) approach is proposed to improve the network lifetime and performance. The proposal addresses both regular traffic scheduling and assignment of the available orthogonal channels in an energy-efficient way. We solve the cooperation between the RL agents problem by using the self-schedule method to accelerate the RL iterations, reduce the communication overhead and balance the energy consumption in the route selection process. Therefore, two algorithms are proposed, the first one is for the Static channel assignment (SSCRL CA) while the second one is for the Dynamic channel assignment (DSCRL CA). The results of extensive simulation experiments show the effectiveness of our approach in improving the network lifetime and performance through the two algorithms.

     

Real-time wrist localization in color images based on corner analysis

Hand detection and gestures recognition have become very popular in recent human-computer interaction systems. Although several methods of hand detection have been proposed in the literature, they exist few methods that use the wrist as a factor of detection, others impose constraints on the length of the sleeves and on the orientation of the hand. In this work, we present a new two-stage algorithm of wrist localization designed for hand detection and gestures recognition systems. The first stage of the algorithm consists in separating the skin region containing the hand from the background, and in the second stage, the wrist is localized from the resulted skin mask. The main contribution of the proposed method is based on the analysis of corners along the contour of the skin masks to localize the wrist emplacement. Based on an evaluation on 437 color images with their ground-truth and three sets of skin masks, we compared our method with other efficient methods of literature and the results obtained were very satisfactory.     

Multi-dimensional Capon spectral estimation using discrete Zhang neural networks

The minimum variance spectral estimator, also known as the Capon spectral estimator, is a high resolution spectral estimator used extensively in practice. In this paper, we derive a novel implementation of a very computationally demanding matched filter-bank based a spectral estimator, namely the multi-dimensional Capon spectral estimator. To avoid the direct computation of the inverse covariance matrix used to estimate the Capon spectrum which can be computationally very expensive, particularly when the dimension of the matrix is large, we propose to use the discrete Zhang neural network for the online covariance matrix inversion. The computational complexity of the proposed algorithm for one-dimensional (1-D), as well as for two-dimensional (2-D) and three-dimensional (3-D) data sequences is lower when a parallel implementation is used.     

Evaluating Regression Models for Temporal Prediction of Wi-Fi Device Mobility

Wireless Personal Communications - The ability to predict the arrival and residence time of mobile users at a particular place is essential for the development of a wealth of new applications and...     

New formulation of GNSS acquisition with CFAR detection

This paper deals with the analysis of the acquisition process performed by a global navigation satellite system (GNSS) receiver with a pilot and data channel or in case of GNSS hybrid receiver. Signal acquisition decides the presence or absence of GNSS signal by comparing signal under test with a fixed threshold and provides a code delay and a Doppler frequency estimation, but in low signal conditions or in a noisy environment; acquisition systems are vulnerable and can give a high false alarm and low detection probability. Firstly, we introduce a cell‐averaging‐constant false alarm rate (CFAR) then a data‐pilot cell‐averaging‐CFAR detector fusion based to deal with these situations. In this context, we use a new mathematical derivation to develop a closed‐form analytic expressions for the probabilities of detection and false alarm. The performances of the proposed detector are evaluated and compared with a non‐CFAR case through an analytical and numerical results validated by Mont Carlo simulations. Copyright © 2016 John Wiley & Sons, Ltd.     

Facial expression recognition using tracked facial actions: Classifier performance analysis

In this paper, we address the analysis and recognition of facial expressions in continuous videos. More precisely, we study classifiers performance that exploit head pose independent temporal facial action parameters. These are provided by an appearance-based 3D face tracker that simultaneously provides the 3D head pose and facial actions. The use of such tracker makes the recognition pose- and texture-independent. Two different schemes are studied. The first scheme adopts a dynamic time warping technique for recognizing expressions where training data are given by temporal signatures associated with different universal facial expressions. The second scheme models temporal signatures associated with facial actions with fixed length feature vectors (observations), and uses some machine learning algorithms in order to recognize the displayed expression. Experiments quantified the performance of different schemes. These were carried out on CMU video sequences and home-made video sequences. The results show that the use of dimension reduction techniques on the extracted time series can improve the classification performance. Moreover, these experiments show that the best recognition rate can be above 90%.     

Designing a camera placement assistance system for human motion capture based on a guided genetic algorithm

Virtual Reality - In multi-camera motion capture systems, determining the optimal camera configuration (camera positions and orientations) is still an unresolved problem. At present, configurations...     

On the self-similarity of 1/f sequences synthesized by recursive filtering

This paper describes an approximate method for synthesizing sequences of statistically self-similar processes and analyses its performance to generate sample sequences with this statistical property. The method is based upon approximating the infinite dimensional difference equation which describes the FARIMA(0, α, 0) model by a finite dimensional difference equation. The parameters estimation for parameterizing the binomial coefficients is performed by using deterministic signal modeling techniques. The three techniques considered are: Prony, Steiglitz MacBride, and Shaw methods. In addition to allow considerable savings in memory requirements and great reduction in computation time, the performance analysis results show that the generated sequences are statistically self-similar in the sense that the estimated Hurst parameter is very close to that imposed in the sequence generator.     

An efficient speech recognition system in adverse conditions using the nonparametric regression

General Regression Neural Networks (GRNN) have been applied to phoneme identification and isolated word recognition in clean speech. In this paper, the authors extended this approach to Arabic spoken word recognition in adverse conditions. In fact, noise robustness is one of the most challenging problems in Automatic Speech Recognition (ASR) and most of the existing recognition methods, which have shown to be highly efficient under noise-free conditions, fail drastically in noisy environments. The proposed system was tested for Arabic digit recognition at different Signal-to-Noise Ratio (SNR) levels and under four noisy conditions: multispeakers babble background, car production hall (factory), military vehicle (leopard tank) and fighter jet cockpit (buccaneer) issued from NOISEX-92 database. The proposed scheme was successfully compared to the similar recognizers based on the Multilayer Perceptrons (MLP), the Elman Recurrent Neural Network (RNN) and the discrete Hidden Markov Model (HMM). The experimental results showed that the use of nonparametric regression with an appropriate smoothing factor (spread) improved the generalization power of the neural network and the global performance of the speech recognizer in noisy environments.     

Robust acoustic bird recognition for habitat monitoring with wireless sensor networks

The key solution to study birds in their natural habitat is the continuous survey using wireless sensors networks (WSN). The final objective of this study is to conceive a system for monitoring threatened bird species using audio sensor nodes. The principal feature for their recognition is their sound. The main limitations encountered with this process are environmental noise and energy consumption in sensor nodes. Over the years, a variety of birdsong classification methods has been introduced, but very few have focused to find an adequate one for WSN. In this paper, a tonal region detector (TRD) using sigmoid function is proposed. This approach for noise power estimation offers flexibility, since the slope and the mean of the sigmoid function can be adapted autonomously for a better trade-off between noise overvaluation and undervaluation. Once the tonal regions in the noisy bird sound are detected, the features gammatone teager energy cepstral coefficients (GTECC) post-processed by quantile-based cepstral normalization were extracted from the above signals for classification using deep neural network classifier. Experimental results for the identification of 36 bird species from Tonga lake (northeast of Algeria) demonstrate that the proposed TRD–GTECC feature is highly effective and performs satisfactorily compared to popular front-ends considered in this study. Moreover, recognition performance, noise immunity and energy consumption are considerably improved after tonal region detection, indicating that it is a very suitable approach for the acoustic bird recognition in complex environments with wireless sensor nodes.