19 research outputs found
Rolling vibes: continuous transport infrastructure monitoring
Get PDFTransport infrastructure is a people to people technology, in the sense that is build by people to serve people, by facilitating transportation, connection and communication. People improved infrastructure by applying simple methods derived from their sensing and thinking. Since the early ages, humans knew that infrastructure should avoid certain amount of vibration that affected either the passenger or the carriage. They achieved that knowledge using the human body sensing capabilities. In fact the human body is the perfect complex sensor node, sensing a wide range of bandwidths, starting from vibrations to the radio waves. Nowadays, specific systems are available that can determine the quality of the road surface by means of special measurements in a somewhat more systematic way. However, a new paradigm of crowd based sensing is emerging, as a result of the smart mobile device revolution. These devices are becoming as ubiquitous as the people that carry them. As daily commuters travel around, their senses are trained to detect road pavement irregularities, uncomfortable turns, potholes, road joints, rail road bumps, stations, accelerations, deceleration and so on. If a human can detect these situations just by classifying the vibration level, why cannot the smartphones in their pocket do the same thing? And same as people do, what a smartphone skipped or missed learning can be compensated by more and more knowledge provided by an army of smartphones participating in infrastructure monitoring. This naive reasoning gave rise to further reasoning and studying the nature of vibration resulting in four long years of research and this thesis. Smartphone based crowd-sensing is not a new concept, there have been previous attempts to make use of this all-around technology. However, a major challenge is the fact that the smartphones are very diverse, have no accurate sensors, and are non-deterministic by nature. On the other hand being ubiquitous, they provide the ability to continuously measure because of the available processor, memory, and (wireless) communication tools. Correct handling of these inaccuracies and unreliable data has been a central theme of the research. Like people, intelligence, learning systems, and additional observations are used to compensate for inaccurate and incomplete observations. The signals measured from the transport infrastructure are a function of four parameters time, distance, temporal frequency and spatial frequency, each with a limited degree of accuracy. In this thesis, we show that with the help of advanced signal processing and machine learning, despite the many inaccuracies in the observations, we can accurately reflect road quality and type of damage. The information obtained by the smartphone sensors is first processed locally by wavelet decomposition methods and useful features are calculated which are then clustered. To compensate for the position inaccuracies, a new aggregation and visualization algorithm has been developed. In addition to a more or less direct measurement of the vibrations, an indirect method is also used, taking into account the driver’s driving behavior. The algorithms, methods, and techniques have been extensively tested and evaluated in various scenarios (for motor vehicles, cyclists, and wheelchairs). Certain indicators are computed to reflect the state-of-the-art requirements. In addition to measuring the quality of the road surface, the quality of railroad track geometry has also been measured
Highly Energy Efficient Animal Mobility Driven BLE Beacon Advertising Control for Wildlife Monitoring
Get PDFWireless sensor networks (WSN) are the basis for wildlife monitoring systems (WMS) because of their flexibility and energy-efficiency. However, their utilization often involves an inherently frequent data advertising mechanism that can result in a significantly higher energy consumption for monitoring mobile animals. To overcome this issue, we propose an mobility event-driven beacon advertising control scheme based on an unsupervised algorithm. We transmit more frequently in a panic state and less frequently during passive state. We compared our approach with a conventional periodic advertising and K-Mean clustering based schemes by analyzing the network energy consumption. The results show that the proposed mechanism could indeed reduce the energy consumption substantially by adaptively adjusting the node wake-up time to specific animal movement states.</p
e-Sight: Real-time cloud platform for visualizing edge transport infrastructure information
Get PDFThe huge amount of streaming information generated by the new wave of edge devices that are used to monitor a plethora of everyday aspects, prompts the need for efficient techniques to handle, process, aggregate, and visualize this stream of data. One such field is the continuous transport infrastructure monitoring, where smartphones inside driving vehicles act as edge sensing and computing nodes to measure the quality of the road pavement among other things. Although the location accuracy of such devices is within the acceptable bounds, the accumulated error can lead to large deviations from the location of interest, reducing the measurement credibility. The data represent a geographically vast infrastructure network in need of real-time, bird-eye visualization. This paper describes the implementation of such a real-time platform and the challenges the task provides. By implementing relatively new open-source cross-platform environments, the platform is scalable and versatile, reducing the costs associated with cloud-based implementation systems
k-SpecNET: Localization and classification of indoor superimposed sound for acoustic sensor networks
Get PDFAutomatic localization and classification of environmental sound events can provide great aid to many human-centric applications. However as many papers have mentioned, environmental sound events in daily life are complicated and hard to classify especially when multiple sounds happen simultaneously. Being different from many other works, we use an acoustic-sensor-network to solve this problem and decompose overlapping sound events using a sound localization model. The core of our contribution is to first find and locate the keypoints from each microphone's spectrogram and then aggregate them. With these aggregated keypoints as input, we then use 2 different classification models to further classify the type of sound sources. Compared with other classification models that only use single microphone, our experiments show that our solution is both accurate and low-cost in terms of calculation effort
SpinSafe: An unsupervised smartphone-based wheelchair path monitoring system
Get PDFMovement and social life of wheelchair users are constrained by their disability and suitability of paths they can move on. Modern electric wheelchairs offer them assisted drive, making their movement easier and longer. They, however, do not prevent accidents, injuries, and inconveniences caused by path roughness and ramp slopes. Providing information about suitability and accessibility of paths and buildings for wheelchair users will enable them to beforehand plan their trip to not to be caught by surprises or not to take a trip all together. The recent emergence of smartphones equipped with inertial sensors offers new opportunities for provision of information regarding quality and accessibility of paths and buildings for wheelchair users. To this end, we propose a smartphone-based participatory system incorporating a hybrid unsupervised machine learning technique based on Self Organized Maps (SOM) to identify path conditions and to create clusters of similar path types. Our solution provides useful information about the angle of the ramp and curb slopes as well as pavement quality and roughness and path types
RoVi: Continuous transport infrastructure monitoring framework for preventive maintenance
Get PDFGround transport infrastructures require in-situ monitoring to evaluate their condition and deterioration and to design appropriate preventive maintenance strategies. Current monitoring practices provide accurate and detailed spatial measurements but often lack the required temporal resolution. This is because the large scale of these infrastructures and the expensive equipments required for monitoring activities do not allow running very frequent measurement campaigns. In this paper, we present RoVi, a novel smartphone-based framework for continuous monitoring of a number of health and condition indicators for variety of ground infrastructures and assets. These indicators include railroad track geometry features such as Cant, Twist, Curvature, and Alignment for different segment lengths as well as road and bike path roughness index (i.e., an equivalent to the International Roughness Index, the so called IRI). RoVi uses an optimized processing algorithm technique on data acquired by smartphones' inertial sensors and relies on sensing, processing power, and networking capabilities of smartphones carried by car/bike drivers and train passengers to provide real time space-time information for fine-grained monitoring of infrastructures. It utilizes the crowd sensing concept to fill in the gap between current sparse consecutive inspections. RoVi provides a reliable and accurate analytic tool for engineers and maintenance planners by offering them features and indicators they require for asset management and maintenance planning. We extract these features and indicators from noisy smartphone data utilizing adaptive signal processing techniques followed by feature calculations and geo-location visualization. Our fast data aggregation algorithm based on Delaunay triangulation updates profiles with new measurements arriving in real time from smartphones. By doing so, it tackles the notorious problem of smartphone GPS accuracy. Performance evaluation of our framework has been performed on measurements collected by smartphones and compared with the ground truth measurements collected by the highend measurement vehicles (i.e., ARAN for roads and UMF120 measurement train for railroads)
Speaker Counting Model based on Transfer Learning from SincNet Bottleneck Layer
Get PDFPeople counting techniques have been widely researched recently and many different types of sensors can be used in this context. In this paper, we propose a system based on a deep-learning model able to identify the number of people in the crowded scenarios through the speech sound. In a nutshell the system relies on two components: counting concurrent speakers in overlapping talking sound directly and clustering single-speaker sound by speaker-identity over time. Compared to previously proposed speaker-counting systems models that only cluster single-speaker sound, this system is more accurate and less vulnerable to the overlapping sound in the crowded environment. In addition, counting speakers in overlapping sound also gives the minimal number of speakers so that it also improves the counting accuracy in a quiet environment. Our methodology is inspired by the newly proposed SincNet deep neural network framework which proves to be outstanding and highly efficient in sound processing with raw signals. By transferring the bottleneck layer of SincNet model as features fed to our speaker clustering model we reached a noticeably better performance than previous models who rely on the use MFCC and other engineered features
Unsupervised learning of wildlife behaviour for activity-driven opportunistic beacon networks
Get PDFMonitoring wild animals in their natural habitat and in real time constitutes an essential aspect of biological and environmental studies. Monitoring is mainly conducted through wireless wildlife monitoring systems (WMS) due to their energy-efficiency and scalability properties. However, using WMS often involves the deployment of energy demanding wireless radio technologies and protocols that significantly increase energy consumption while tracking mobile animals. Thanks to the raise of IoT devices capable of sensing, computing, and wireless networking, WMS can become more efficient and overcome the initial drawbacks. This paper, describes an activity driven beaconing mechanism based on unsupervised activity classification scheme. The algorithm is evaluated for different parameters involving the sampling rate, processing window as well as different cluster sizes. The evaluation shows that use of lightweight algorithms and low sampling rates provides the possibility to reliably monitor the activity of the animal. The evaluation results showed that the proposed mechanism could reduce energy consumption by increasing communication sleep-time while the objects were stationary
A sound-based crowd activity recognition with neural network based regression models
Get PDFActivities performed by humans can be recognized by the sound they emit while being performed, hence, researchers have proposed methods that use sound to recognize human activities, by detecting the presence of sound events in short time frames. However, in crowded environments, many sound events overlap making it impossible to distinguish the individual events and methods of detection to fail.To address this issue and make the sound-based model suitable for crowd activities, this paper proposes to predict the proportion of activities happening in a specific place, by designing two neural network-based regression models: a CNN-model and a concatenate model. The CNN-model takes the Mel-bands as the input and is very popular in single activity recognition problems. Based on the CNN-model, we also designed a concatenate model which additionally inputting the global FFT feature to further improve the performance.The evaluation of this approach is performed over 3 generated groups of audio samples, where each group has a different crowded-level. Both RMSE and coefficient of determination (R2 score), are used as evaluation metrics. The experiments show that the concatenate model works statistically better throughout the dataset, with a R2 score of 0.7377. Results show that using the concatenate model with both short-frame and holistic features provides a better result than any single-feature based model
