Patterns In Sensor Data Research Articles

Tracking changes in user activity from unlabelled smart home sensor data using unsupervised learning methods

This paper investigates the utility of unsupervised machine learning and data visualisation for tracking changes in user activity over time. This is done through analysing unlabelled data generated from passive and ambient smart home sensors, such as motion sensors, which are considered less intrusive than video cameras or wearables. The challenge in using unlabelled passive and ambient sensors data for activity recognition is to find practical methods that can provide meaningful information to support timely interventions based on changing user needs, without the overhead of having to label the data over long periods of time. The paper addresses this challenge to discover patterns in unlabelled sensor data using kernel density estimation (KDE) for pre-processing the data, together with t-distributed stochastic neighbour embedding and uniform manifold approximation and projection for visualising changes. The methodology is developed and tested on the Aruba CASAS smart home dataset and focusses on discovering and tracking changes in kitchen-based activities. The traditional approach of using sliding windows to segment the data requires a priori knowledge of the temporal characteristics of activities being identified. In this paper, we show how an adaptive approach for segmentation, KDE, is a suitable alternative for identifying temporal clusters of sensor events from unlabelled data that can represent an activity. The ability to visualise different recurring patterns of activity and changes to these over time is illustrated by mapping the data for separate days of the week. The paper then demonstrates how this can be used to track patterns over longer time-frames which could be used to help highlight differences in the user’s day-to-day behaviour. By presenting the data in a format that can be visually reviewed for temporal changes in activity over varying periods of time from unlabelled sensor data, opens up the opportunity for carers to then initiate further enquiry if variations to previous patterns are noted. This is seen as an accessible first step to enable carers to initiate informed discussions with the service user to understand what may be causing these changes and suggest appropriate interventions if the change is found to be detrimental to their well-being.

Unsupervised Factory Activity Recognition with Wearable Sensors Using Process Instruction Information

This paper presents an unsupervised method for recognizing assembly work done by factory workers by using wearable sensor data. Such assembly work is a common part of line production systems and typically involves the factory workers performing a repetitive work process made up of a sequence of manual operations, such as setting a board on a workbench and screwing parts onto the board. This study aims to recognize the starting and ending times for individual operations in such work processes through analysis of sensor data collected from the workers along with analysis of the process instructions that detail and describe the flow of operations for each work process. We propose a particle-filter-based factory activity recognition method that leverages (i) trend changes in the sensor data detected by a nonparametric Bayesian hidden Markov model, (ii) semantic similarities between operations discovered in the process instructions, (iii) sensor-data similarities between consecutive repetitions of individual operations, and (iv) frequent sensor-data patterns (motifs) discovered in the overall assembly work processes. We evaluated the proposed method using sensor data from six workers collected in actual factories, achieving a recognition accuracy of 80% (macro-averaged F-measure).

Wireless Sensor Network System for Landslide Monitoring and Warning

This paper presents a wireless sensor network system (WSNS) for effective, reliable, and efficient monitoring of landslides. The system incorporates a network of wireless inertial measurement unit (IMU) sensor devices for collecting movement data, a local base station for data gathering, a capture server for data processing and storage, and a warning system. The major contributions of this paper are: 1) two approaches for defining movement thresholds; 2) landslide classification concept based on IMU sensor data patterns and magnitudes; and 3) a conceptual framework for building an intelligent and reliable wireless monitoring and warning system. The IMU sensor data collected by three-axis accelerometer and three-axis gyroscope were utilized to define the movement thresholds and classify landslides based on specially designed laboratory experiments. The performances of the IMU sensors and the base station for data collection and communication were tested through a rock-fall experiment conducted in the field conditions. The WSN-IMU system is capable of monitoring all types of slope movements independent of the triggering factors. The unique ability of the WSN-IMU system to determine landslide types allows designers and authorized personnel to predict subsequent movement pattern and duration so as to implement appropriate risk management and control measures to alleviate the socioeconomic losses. This paper outcome serves as the foundation for future studies and technological advancements that will facilitate landslide stabilization or mitigation actions as well as to predict the intensity of damages associated with those landslides.

Randomized Trial of Intelligent Sensor System for Early Illness Alerts in Senior Housing

ObjectivesMeasure the clinical effectiveness and cost effectiveness of using sensor data from an environmentally embedded sensor system for early illness recognition. This sensor system has demonstrated in pilot studies to detect changes in function and in chronic diseases or acute illnesses on average 10 days to 2 weeks before usual assessment methods or self-reports of illness. DesignProspective intervention study in 13 assisted living (AL) communities of 171 residents randomly assigned to intervention (n=86) or comparison group (n=85) receiving usual care. MethodsIntervention participants lived with the sensor system an average of one year. MeasurementsContinuous data collected 24 hours/7 days a week from motion sensors to measure overall activity, an under mattress bed sensor to capture respiration, pulse, and restlessness as people sleep, and a gait sensor that continuously measures gait speed, stride length and time, and automatically assess for increasing fall risk as the person walks around the apartment. Continuously running computer algorithms are applied to the sensor data and send health alerts to staff when there are changes in sensor data patterns. ResultsThe randomized comparison group functionally declined more rapidly than the intervention group. Walking speed and several measures from GaitRite, velocity, step length left and right, stride length left and right, and the fall risk measure of functional ambulation profile (FAP) all had clinically significant changes. The walking speed increase (worse) and velocity decline (worse) of 0.073 m/s for comparison group exceeded 0.05 m/s, a value considered to be a minimum clinically important difference. No differences were measured in health care costs. ConclusionsThese findings demonstrate that sensor data with health alerts and fall alerts sent to AL nursing staff can be an effective strategy to detect and intervene in early signs of illness or functional decline.

Sequential Behavior Pattern Discovery with Frequent Episode Mining and Wireless Sensor Network

By recognizing patterns in occupants' daily activities, building systems are able to optimize and personalize services. Established technologies are available for data collection and pattern mining, but they all share the drawback that the methodology used for data collection tends to be ill suited for pattern recognition. For this research, we developed a bespoke WSN and combined it with a compact data format for frequent episode mining to overcome this obstacle. The proposed framework has been evaluated with both synthetic data from a smart home simulator and with real data from a self-organizing WSN in a student's home. We are able to demonstrate that the framework is capable of discovering sequential patterns in heterogeneous sensor data. With corresponding scenarios, patterns in daily activities can be deduced. The framework is self-contained, scalable, and energy-efficient, and is thus applicable in multiple building system settings.

Spatio-temporal Pattern discovery in sensor data: A multivalued decision systems approach

Discovering novel and interesting spatio-temporal patterns in sensor data is a major challenge in many scientific domains. Such data are often continuous, unbounded, and associated with high speed, time-variant distribution with local and spatial trends. This paper presents a formalism that includes an extension of classical rough set inference mechanism to reason with space-time variant data streams. The concept of multivalued decision systems has been used to account for multiple patterns in a single time window or snapshot. Such patterns or templates are incorporated in rough set-based rule induction process. A framework for sensor data integration is illustrated by using a space-time clustering mechanism followed by the generation of templates and local rules from such clusters. The multivalued decision system allows mining complex multiple patterns instead of a single value in a given template without requiring complex feature transformation. It also allows us to quantify and estimate potential data compression and associated uncertainty parameters. Finally, the results are validated and compared with other related algorithms. In general, the framework will help us understand the underlying reasoning mechanism about the “part and whole” or spatio-temporal mereological relationship without sacrificing the semantics of the sensor data attributes.

Bayesian aggregation of evidence for detection and characterization of patterns in multiple noisy observations

Bayesianism is a premier representation for encoding intelligence in robotic sensing. My thesis develops Bayesian Aggregation (BA), a machine learning pipeline for understanding patterns in noisy sensor data and extracting structure from multiple correlated observations.

Investigation of an expert health monitoring system for aeronautical structures based on pattern recognition and acousto-ultrasonics

Aeronautical structures are subjected to damage during their service raising the necessity for periodic inspection and maintenance of their components so that structural integrity and safe operation can be guaranteed. Cost reduction related to minimizing the out-of-service time of the aircraft, together with the advantages offered by real-time and safe-life service monitoring, have led to a boom in the design of inexpensive and structurally integrated transducer networks comprising actuators, sensors, signal processing units and controllers. These kinds of automated systems are normally referred to as smart structures and offer a multitude of new solutions to engineering problems and multi-functional capabilities. It is thus expected that structural health monitoring (SHM) systems will become one of the leading technologies for assessing and assuring the structural integrity of future aircraft. This study is devoted to the development and experimental investigation of an SHM methodology for the detection of damage in real scale complex aeronautical structures. The work focuses on each aspect of the SHM system and highlights the potentialities of the health monitoring technique based on acousto-ultrasonics and data-driven modelling within the concepts of sensor data fusion, feature extraction and pattern recognition. The methodology is experimentally demonstrated on an aircraft skin panel and fuselage panel for which several damage scenarios are analysed. The detection performance in both structures is quantified and presented.

Critical analysis of smart environment sensor data behavior pattern based on sequential data mining techniques

Purpose – Behavioral pattern mining for intelligent system such as SmEs sensor data are vitally important in many applications and performance optimizations. Sensor pattern mining (SPM) is also dynamic and a hot research issue to pervasive and ubiquitous of smart technologies toward improving human life. However, in large-scale sensor data, exploring and mining pattern, which leads to detect the abnormal behavior is challenging. The paper aims to discuss these issues. Design/methodology/approach – Sensor data are complex and multivariate, for example, which data captured by the sensors, how it is precise, what properties are recorded or measured, are important research issues. Therefore, the method, the authors proposed Sequential Data Mining (SDM) approach to explore pattern behaviors toward detecting abnormal patterns for smart space fault diagnosis and performance optimization in the intelligent world. Sensor data types, modeling, descriptions and SPM techniques are discussed in depth using real sensor data sets. Findings – The outcome of the paper is measured as introducing a novel idea how SDM technique’s scale-up to sensor data pattern mining. In the paper, the approach and technicality of the sensor data pattern analyzed, and finally the pattern behaviors detected or segmented as normal and abnormal patterns. Originality/value – The paper is focussed on sensor data behavioral patterns for fault diagnosis and performance optimizations. It is other ways of knowledge extraction from the anomaly of sensor data (observation records), which is pertinent to adopt in many intelligent systems applications, including safety and security, efficiency, and other advantages as the consideration of the real-world problems.

Multi-objective monitoring of closed-loop controlled systems using adaptive LASSO

Statistical process control detects nonrandom deviations from a normal distribution. However, most industrial processes are equipped with feedback control loops to reject such disturbances. Hence it is the failure or performance degradation that we should detect instead of mere sensor data patterns that show deviations from normal. In this paper, a multi-objective monitoring approach is proposed to monitor both the system stability and performance. First, closed-loop output data are fitted with an autoregressive moving average model with exogenous inputs (ARMA(X)), using adaptive least absolute shrinkage and selection operator (LASSO). The system stability can be monitored using the largest absolute value of the roots of AR terms. The performance of feedback controller can be monitored based on the Minimum-Variance (MV) principles. The proposed approach was applied to several simulation examples to show that the above two monitoring objects are only sensitive to the faults in processes, but not to disturbances that can be handled by the controller. The ability of the method to localized control loop failure was demonstrated using the Tennessee–Eastman benchmark problem. In addition, we found that the method was able to identify the failure of the outer loop while the inner loop is still performing in a cascade control.

A Multidimensional Time-Series Similarity Measure With Applications to Eldercare Monitoring.

In the last decade, data mining techniques have been applied to sensor data in a wide range of application domains, such as healthcare monitoring systems, manufacturing processes, intrusion detection, database management, and others. Many data mining techniques are based on computing the similarity between two sensor data patterns. A variety of representations and similarity measures for multiattribute time series have been proposed in the literature. In this paper, we describe a novel method for computing the similarity of two multiattribute time series based on a temporal version of Smith-Waterman (SW), a well-known bioinformatics algorithm. We then apply our method to sensor data from an eldercare application for early illness detection. Our method mitigates difficulties related to data uncertainty and aggregation that often arise when processing sensor data. The experiments take place at an aging-in-place facility, TigerPlace, located in Columbia, MO, USA. To validate our method, we used data from nonwearable sensor networks placed in TigerPlace apartments, combined with information from an electronic health record. We provide a set of experiments that investigate temporal version of SW properties, together with experiments on TigerPlace datasets. On a pilot sensor dataset from nine residents, with a total of 1902days and around 2.1 million sensor hits of collected data, we obtained an average abnormal events prediction F-measure of 0.75.

Data-Driven Rule Mining and Representation of Temporal Patterns in Physiological Sensor Data.

Mining and representation of qualitative patterns is a growing field in sensor data analytics. This paper leverages from rule mining techniques to extract and represent temporal relation of prototypical patterns in clinical data streams. The approach is fully data-driven, where the temporal rules are mined from physiological time series such as heart rate, respiration rate, and blood pressure. To validate the rules, a novel similarity method is introduced, that compares the similarity between rule sets. An additional aspect of the proposed approach has been to utilize natural language generation techniques to represent the temporal relations between patterns. In this study, the sensor data in the MIMIC online database was used for evaluation, in which the mined temporal rules as they relate to various clinical conditions (respiratory failure, angina, sepsis, …) were made explicit as a textual representation. Furthermore, it was shown that the extracted rule set for any particular clinical condition was distinct from other clinical conditions.

A Bioinspired Methodology Based on an Artificial Immune System for Damage Detection in Structural Health Monitoring

Among all the aspects that are linked to a structural health monitoring (SHM) system, algorithms, strategies, or methods for damage detection are currently playing an important role in improving the operational reliability of critical structures in several industrial sectors. This paper introduces a bioinspired strategy for the detection of structural changes using an artificial immune system (AIS) and a statistical data-driven modeling approach by means of a distributed piezoelectric active sensor network at different actuation phases. Damage detection and classification of structural changes using ultrasonic signals are traditionally performed using methods based on the time of flight. The approach followed in this paper is a data-based approach based on AIS, where sensor data fusion, feature extraction, and pattern recognition are evaluated. One of the key advantages of the proposed methodology is that the need to develop and validate a mathematical model is eliminated. The proposed methodology is applied, tested, and validated with data collected from two sections of an aircraft skin panel. The results show that the presented methodology is able to accurately detect damage.

Behavior analysis of elderly using topic models

This paper describes two new topic models for the analysis of human behavior in homes that are equipped with sensor networks. The models are based on Latent Dirichlet Allocation (LDA) topic models and can detect patterns in sensor data in an unsupervised manner. LDA–Gaussian, the first variation of the model, is a combination of a Gaussian Mixture Model and the LDA model. Here the multinomial distribution that is normally used in the LDA model is replaced by a set of Gaussian distributions. LDA–Poisson, the second variation of the model, uses a set of Poisson distribution to model the observations. The Poisson distribution is better suited to handle counts of stochastic events but less well-suited to model time. For this we use the von Mises distribution, resulting in ‘LDA–Poisson–von-Mises’. The parameters of the models are determined with an EM-algorithm. The models are evaluated on more than 450 days of real-world sensor data, gathered in the homes of five elderly people, and are compared with a baseline approach where standard k-means clustering is used to quantize the data. We show that the new models find more meaningful topics than the baseline and that a semantic description of these topics can be given. We also evaluated the models quantitatively, using perplexity as measure for the model fit. Both LDA–Gaussian and LDA–Poisson result in much better models than the baseline, and our experiments show that, of the proposed models, the LDA–Poisson–von-Mises model performs best.

Fundamental Considerations: Impact of Sensor Characteristics, Application Environments in Wireless Sensor Networks

Observed from the recent performance evaluation of clustering schemes in wireless sensor networks, we found that most of them did not consider various sensor characteristics and its application environment. Without considering these, the performance evaluation results are difficult to be trusted because these networks are application-specific. In this paper, for the fair evaluation, we measured several clustering scheme's performance variations in accordance with sensor data pattern, number of sensors per node, density of points of interest (data density) and sensor coverage. According to the experiment result, we can conclude that clustering methods are easily influenced by POI variation. Network lifetime and data accuracy are also slightly influenced by sensor coverage and number of sensors. Therefore, in the case of the clustering scheme that did not consider various conditions, fair evaluation cannot be expected.

Applying additive logistic regression to data derived from sensors monitoring behavioral and physiological characteristics of dairy cows to detect lameness

The hypothesis was that sensors currently available on farm that monitor behavioral and physiological characteristics have potential for the detection of lameness in dairy cows. This was tested by applying additive logistic regression to variables derived from sensor data. Data were collected between November 2010 and June 2012 on 5 commercial pasture-based dairy farms. Sensor data from weigh scales (liveweight), pedometers (activity), and milk meters (milking order, unadjusted and adjusted milk yield in the first 2min of milking, total milk yield, and milking duration) were collected at every milking from 4,904 cows. Lameness events were recorded by farmers who were trained in detecting lameness before the study commenced. A total of 318 lameness events affecting 292 cows were available for statistical analyses. For each lameness event, the lame cow’s sensor data for a time period of 14d before observation date were randomly matched by farm and date to 10 healthy cows (i.e., cows that were not lame and had no other health event recorded for the matched time period). Sensor data relating to the 14-d time periods were used for developing univariable (using one source of sensor data) and multivariable (using multiple sources of sensor data) models. Model development involved the use of additive logistic regression by applying the LogitBoost algorithm with a regression tree as base learner. The model’s output was a probability estimate for lameness, given the sensor data collected during the 14-d time period. Models were validated using leave-one-farm-out cross-validation and, as a result of this validation, each cow in the data set (318 lame and 3,180 nonlame cows) received a probability estimate for lameness. Based on the area under the curve (AUC), results indicated that univariable models had low predictive potential, with the highest AUC values found for liveweight (AUC=0.66), activity (AUC=0.60), and milking order (AUC=0.65). Combining these 3 sensors improved AUC to 0.74. Detection performance of this combined model varied between farms but it consistently and significantly outperformed univariable models across farms at a fixed specificity of 80%. Still, detection performance was not high enough to be implemented in practice on large, pasture-based dairy farms. Future research may improve performance by developing variables based on sensor data of liveweight, activity, and milking order, but that better describe changes in sensor data patterns when cows go lame.

Data Pattern With ECT Sensor and Its Impact on Image Reconstruction

Electrical capacitance tomography (ECT) is a tomographic imaging technique for monitoring and studying dielectric processes. Image reconstruction in ECT is a nonlinear and typical inverse problem and has been widely investigated. In this paper, we investigate the ECT sensor data pattern and its impact on image reconstruction. By analyzing the capacitance data of certain typical permittivity distributions, the non-linearity with ECT image reconstruction is characterized. A modified linear model for ECT image reconstruction is presented, in which the contributions to image reconstruction due to the capacitance between different electrode pairs are considered. Additionally, the error associated with ECT sensor sensitivity distribution is taken into account. Based on the presented model, a total least squares and total variation based regularization method is used to reconstruct the permittivity distribution. The results based on both simulation and experimental data show that the presented model together with the total least squares and total variation based regularization method can produce fairly high quality reconstructions with clear edges and high precision.

Impact of Sensor Data Patterns on Performance Evaluation of Clustering Schemes in Wireless Sensor Networks

Impact of Sensor Data Patterns on Performance Evaluation of Clustering Schemes in Wireless Sensor Networks

On retrieving patterns in environmental sensor data

As many sensor networks are currently being deployed for environmental monitoring, there is a growing need to develop systems and applications for managing, processing and retrieving massive amounts of data generated from those networks. In this research, a query answering system with pattern mining techniques is investigated specifically for marine sensor data. We consider three applications of pattern mining: similar pattern search, predictive query and query by clustering. In pattern mining for query answering, we adopt the dynamic time warping (DTW) method for similarity measurement. We also propose the use of a query relaxation approach that recommends users change parameters of a given query to get an answer. Finally, we show implementation results of pattern query answering in a marine sensor network deployed in the South East of Tasmania, Australia. Pattern query answering system benefits in accessing and discovering knowledge from sensor data for decision making purposes.

Ubicorder: A Mobile Device for Situated Interactions With Sensor Networks

The Ubicorder is a mobile, location and orientation aware device for browsing and interacting with real-time sensor network data. In addition to browsing data, the Ubicorder also provides a graphical user interface (GUI) that users can use to define inference rules. These inference rules detect sensor data patterns, and translate them to higher order events. Rules can also be recursively combined to form an expressive and robust vocabulary for detecting real-world phenomena, thus enabling users to script higher level and relevant responses to distributed sensor stimuli. The Ubicorder's mobile, handheld form-factor enables users to easily bring the device to the phenomena of interest, hence simultaneously observe or cause real-world stimuli and manipulate in-situ the event detection rules easily using its graphical interface. In a first-use user study, participants without any prior sensor network experience rated the Ubicorder highly for its usefulness and usability when interacting with a sensor network.