Majid Bahrepour

Sensor fusion-based event detection in Wireless Sensor Networks

Recently, Wireless Sensor Networks (WSN) community has witnessed an application focus shift. Although, monitoring was the initial application of wireless sensor networks, in-network data processing and (near) real-time actuation capability have made wireless sensor networks suitable candidate for event detection and alarming applications as well. Unreliability and dynamic (e.g. in terms of deployment area, network resources, and topology) are normal practices in the field of WSN. Therefore, effective and trustworthy event detection techniques for the WSN require robust and intelligent methods of mining hidden patterns in the sensor data, while supporting various kinds of dynamicity. Due to the fact that events are often functions of more than one attribute, data fusion and use of more features can help increasing event detection rate and reducing false alarm rate. In addition, sensor fusion can lead to more accurate and robust event detection by eliminating outliers and erroneous readings of individual sensor nodes and combining individual event detection decisions. In this paper, we propose a two-level sensor fusion-based event detection technique for the WSN. The first level of event detection in our proposed approach is conducted locally inside the sensor nodes, while the second level is carried out in a level higher (e.g., in a cluster head or gateway) and incorporates a fusion algorithm to reach a consensus among individual detection decisions made by sensor nodes. By considering fire as an event, we evaluate our approach through several experiments and illustrate impact of sensor fusion on achieving better results.

Use of event detection approaches for outlier detection in wireless sensor networks

Outliers or anomalies are generally considered to be those observations that are considerably diverged from normal pattern of data. Due to their special characteristics, e.g. constrained available resources, frequent physical failure, and often harsh deployment area, wireless sensor networks (WSNs) are more likely to generate outliers compared to their other wireless counterparts. Potential sources of deviated data in a series of measurements are errors, events, and/or malicious attacks on the network. Current studies tend to handle events and errors separately and propose different techniques for event detection as for outlier detection. By bringing the concept of outlier and event close together and assuming that events are some sorts of outliers, in this paper, we investigate applicability of pattern matching-based event detection techniques for outlier detection. Through extensive experiments, we evaluate performance of various event detection techniques to detect outliers and compare them with a recent outlier detection study.

Use of wireless sensor networks for distributed event detection in disaster management applications

Recently, wireless sensor networks (WSNs) have become mature enough to go beyond being simple fine-grained continuous monitoring platforms and have become one of the enabling technologies for early-warning disaster systems. Event detection functionality of WSNs can be of great help and importance for (near) real-time detection of, for example, meteorological natural hazards and wild and residential fires. From the data-mining perspective, many real world events exhibit specific patterns, which can be detected by applying machine learning (ML) techniques. In this paper, we introduce ML techniques for distributed event detection in WSNs and evaluate their performance and applicability for early detection of disasters, specifically residential fires. To this end, we present a distributed event detection approach incorporating a novel reputation-based voting and the decision tree and evaluate its performance in terms of detection accuracy and time complexity.

Fire data analysis and feature reduction using computational intelligence methods

Fire is basically the fast oxidation of a substance that produces gases and chemical productions. These chemical productions can be read by sensors to yield an insight about type and place of the fire. However, as fires may occur in indoor or outdoor areas, the type of gases and therefore sensor readings become different. Recently, wireless sensor networks (WSNs) have been used for environmental monitoring and real-time event detection because of their low implementation costs and their capability of distributed sensing and processing. In this paper, the authors investigate spatial analysis of data for indoor and outdoor fires using data-mining approaches for WSN-based fire detection purposes. This paper also delves into correlated data features in fire data sets and investigates the most contributing features for fire detection applications.

Sensor Fusion-based Activity Recognition for Parkinson Patients

Parkinson disease (PD) is a slow destructive disorder of the central nervous system in which dopamine, i.e., catecholamine neurotransmitter in the central nervous system is lost. PD hurts patients’ movement and speech ability. Sometimes, it can also affect patients’ mood, behavior, and thinking ability. Falling down is a common problem in PD patients and on time fall detection is important to assist PD patients and prevent them from being injured. To this end, being able to correctly distinguish various activities, e.g. walking, sitting, standing still, is a must. To monitor activities and moving patterns of PD patients, a wireless body sensor network (BSN) may prove to be useful. By attaching various wireless sensor nodes on the body of PD patients or integrating them into their shoes or cloths, their activities and physiological conditions can be checked regularly and an alarm can be generated in case of emergency or need for additional assistances. A wireless body sensor network consists of a number of wireless sensor nodes that cooperatively monitor physical (e.g. motion) and physiological (e.g. heart rate) conditions of a person. In addition to sensors, each sensor node is typically equipped with a radio transceiver or other wireless communication devices, a small microcontroller as processing unit, and an energy source in a form of a battery. Sensor nodes may vary in size and type of sensors they are equipped with. Size and cost constraints on sensor nodes cause limitations on their resources in terms of energy, memory, and computational processing. Figure 1 shows an example of a body sensor network. Previous studies for activity recognition of PD patients mostly use accelerometer and occasionally gyroscope sensors attached to various parts of patients’ body (JJ., HA. et al. 1991; Aminian, Robert et al. 1999; JI., AA. et al. 2001; JI., V. et al. 2004; N., T. et al. 2004; White, Wagenaar et al. 2006; Moorea, MacDougalla et al. 2007; Salarian, Russmann et al. 2007). One of the main criticisms on the previous studies is that they use centralized techniques which not only require expensive equipments to monitor physiological conditions and activities of patients [e.g. Vitaport 3 (White, Wagenaar et al. 2006)] but also introduce delays in the detection process. Also due to having a single point of failure they are more prone to failures and crashes. In contrary, we propose a fusion-based distributed algorithm which can be easily implemented on resource constrained wireless sensor nodes and detect and distinguish activities in (near) real-time. Our approach offers three main advantages: (i) distributed processing and reasoning which decreases the data processing

On the Effects of Input Unreliability on Classifion Algorithms

The abundance of data available on Wireless Sensor Networks makes online processing necessary. In industrial applications, for example, the correct operation of equipment can be the point of interest. The raw sampled data is of minor importance. Classification algorithms can be used to make state classifications based on the available data for devices such as industrial refrigerators. The reliability through redundancy approach used in Wireless Sensor Networks complicates practical realizations of classification algorithms. Individual inputs are susceptible to multiple disturbances like hardware failure, communication failure and battery depletion. In order to demonstrate the effects of input failure on classification algorithms, we have compared three widely used algorithms in multiple error scenarios. The compared algorithms are Feed Forward Neural Networks, naive Bayes classifiers and decision trees. Using a new experimental data-set, we show that the performance under error scenarios degrades less for the naive Bayes classifier than for the two other algorithms.

Online unsupervised event detection in wireless sensor networks

Event detection applications of wireless sensor networks (WSNs) highly rely on accurate and timely detection of out of ordinary situations. Majority of the existing event detection techniques designed for WSNs have focused on detection of events with known patterns requiring a priori knowledge about events being detected. In this paper, however, we propose an online unsupervised event detection technique for detection of unknown events. Traditional unsupervised learning techniques cannot directly be applied in WSNs due to their high computational and memory complexities. To this end, by considering specific resource limitations of the WSNs we modify the standard K-means algorithm in this paper and explore its applicability for online and fast event detection in WSNs. For performance evaluation, we investigate event detection accuracy, false alarm, similarity calculation (using the Rand Index), computational and memory complexity of the proposed approach on two real datasets.

Can wireless sensor networks be emotional? A survey of computational models of emotions and their applications for wireless sensor networks

Advances in psychology have revealed that emotions and rationality are interlinked and emotions are essential for rational behaviour and decision making. Therefore, integration of emotions with intelligent systems has become an important topic in engineering. The integration of emotions into intelligent systems requires computational models to generate emotions from external and internal sources. This paper first provides a survey of current computational models of emotion and their applications in engineering. Finally, it assesses potential of integrating emotions in wireless sensor networks (WSNs) by listing some use scenarios and by giving one model application. In this model application performance of a neural network for event detection has been improved using brain emotional learning based intelligent controller (BELBIC).