Human Activity Recognition

Gait recognition is a technique that identifies or verifies people based upon their walking patterns. Smartwatches, which contain an accelerometer and gyroscope have recently been used to implement gait-based biometrics. However, this prior work relied upon data from single sessions for both training and testing, which is not realistic and can lead to overly optimistic performance results. This paper aims to remedy some of these problems by training and evaluating a smartwatch-based biometric system on data obtained from different days. Also, it proposes an advanced feature selection approach to identify optimal features for each user. Two experiments are presented under three different scenarios: Same-Day, Mixed-Day, and Cross-Day. Competitive results were achieved (best EERs of 0.13% and 3.12% by using the Same day data for accelerometer and gyroscope respectively and 0.69% and 7.97% for the same sensors under the Cross-Day evaluation. The results show that the technology is sufficiently capable and the signals captured sufficiently discriminative to be useful in performing gait recognition.

The increasing popularity of smart wearable devices can be attributed to progress in research towards human activity recognition. With the use of tiny sensing units, the different human activities of day-today life can be identified with a high accuracy. However, as a fitness tracking product, smart wearables are not always accurate in determining actual physical motion details. For example, recognizing walk activity may not always be possible when the monitoring device is held in hand, or kept in pockets. For precise recognition of walk activity, movement of legs needs to be monitored. However, the movement of legs while walking must be distinguished from simple leg swing activities. The present work designs an IMU based sensing system that can prevent false identification of a mimicked walk or leg swing in sitting posture as a real walk activity, using conventional and convolutional deep learning algorithms. The system shows remarkable capability of identifying an actual walk from a mimicked walk activity using CNN 95% of the time.

Human activity monitoring is one of those research areas whose importance and popularity have notably increased in recent years. The popularity of this topic increased in the previous years. Most of the used movement analysis techniques in the area are based on the measurement of the acceleration change of different parts of the body. This is done by attaching one or more little devices with an accelerometer to the body of the observed patient.
Usually, the role of the body-attached devices are only data acquisition, the processing of the acquired data happens offline. This article presents a new solution for this task which combines digital time-frequency signal processing with a parallel programming approach.

Human Activity Recognition(HAR) is classifying activity of a person using responsive sensors that are affected from human movement. Both users and capabilities(sensors) of smartphones increase and users usually carry their smartphone with them. These facts makes HAR more important and popular. This work focuses on recognition of human activity using smartphone sensors using different machine learning classification approaches. Data retrieved from smart phones' accelerometer and gyroscope sensors are classified in order to recognise human activity. Results of the approaches used are compared in terms of efficiency and presicion.

Human Activity Recognition (HAR) has become a subject of high interest within many fields. In this work, the target is to build a classifier with high accuracy, so two approaches are used to build a classifier. The first approach represented by model (A), and the second represented by the model (B), both models are built using four machine learning classification algorithms, which are Random Forest (RF), Naïve Bayes (NB), Support Vector Machines (SVM), and Multilayer perceptron (MLP). Model (A) depends on a single classifier for recognizing 12 activities and Model (B) depends on applying the ensemble technique and build in two steps for recognizing the same activities. The experimental comparison result shows that model (B) improved the accuracy of model (A), as the accuracy is increased with 3.6% in Random Forest, 4.7% in Multilayer perceptron, and 23.2% in Naïve Bayes but for Support Vector Machines, model (A) achieved higher accuracy than model (B).

Human Action Recognition (HAR) is an application-oriented field that utilizes numerous Machine Learning methods to identify diverse human actions or movements to provide an appropriate or suitable response. A HAR method's success largely depends on the performance of the algorithms working in the background for data processing and activity prediction. This article inscribes an easy action recognition method based on a sequential Deep Learning model. Instead of using intricated signal processing, feature extraction, and feature selection techniques, we used a special type of matrix formulated directly from the raw data. The described method has been tested on three different HAR datasetsnamely UCI HAR, WISDM, and KU-HARand the maximum classification accuracies achieved on them were 98.99%, 96.53%, and 96.67%, respectively. Other evaluation metrics yield similar outcomes, which asserts the method's reliability. Detailed descriptions of each step of the methodology, the reported results, and limitations have been provided with necessary illustrations.

Smartwatches, which contain an accelerometer and gyroscope, have recently been used to implement gait/activity-based biometrics. However, many research questions have not been addressed in the prior work such as the training and test data was collected in the same day from a limited dataset, using unrealistic activities (e.g., punch) and/or the authors did not carry out any particular study to identify the most discriminative features. This paper aims to highlight the impact of these factors on the biometric performance. The acceleration and gyroscope data of the gait and game activity was captured from 60 users over multiple days, which resulted in a totally of 24 h of the user's movement. Segment-based approach was used to divide the time-series acceleration and gyroscope data. When the cross-day evaluation was applied, the best obtained EER was 0.69%, and 4.54% for the walking and game activities respectively. The EERs were significantly reduced into 0.05% and 2.35% for the above activities by introducing the majority voting schema. These results were obtained by utilizing a novel feature selection process in which the system minimizing the number of features and maximizing the discriminative information. The results have shown that smartwatch-based activity recognition has significant potential to recognize individuals in a continuous and user friendly approach.

We are currently using smart phone sensors to detect physical activities. The sensors which are currently being used are accelerometer, gyroscope, barometer, etc. Recently, smart phones, equipped with a rich set of sensors, are explored as alternative platforms for human activity recognition. Automatic recognition of physical activities – commonly referred to as human activity recognition (HAR) – has emerged as a key research area in human-computer interaction (HCI) and mobile and ubiquitous computing. One goal of activity recognition is to provide information on a user’s behavior that allows computing systems to proactively assist users with their tasks. Human activity recognition requires running classification algorithms, originating from statistical machine learning techniques. Mostly, supervised or semi-supervised learning techniques are utilized and such techniques rely on labeled data, i.e., associated with a specific class or activity. In most of the cases, the user is required to label the activities and this, in turn, increases the burden on the user. Hence, user- independent training and activity recognition are required to foster the use of human activity recognition systems where the system can use the training data from other users in classifying the activities of a new subject.

Currently, the health care problem for older people is becoming of considerable size and involves the whole civilized world. The number of people in need of assistance is constantly growing due to life extension, and the social cost is becoming unsustainable. Technological evolution and IT development can help contain spending by providing efficient health services and improving the quality of life. This article highlights methodological and technical aspects to support the development of an integrated system for locating and monitoring people in their homes or nursing homes. This system complements the home care service by avoiding hospitalization in health facilities with more significant financial burdens.

During the last 5 years, research on Human Activity Recognition (HAR) has reported on systems showing good overall recognition performance. As a consequence, HAR has been considered as a potential technology for e-health systems. Here, we propose a machine learning based HAR classifier. We also provide a full experimental description that contains the HAR wearable devices setup and a public domain dataset comprising 165,633 samples. We consider 5 activity classes, gathered from 4 subjects wearing accelerometers mounted on their waist, left thigh, right arm, and right ankle. As basic input features to our classifier we use 12 attributes derived from a time window of 150ms. Finally, the classifier uses a committee AdaBoost that combines ten Decision Trees. The observed classifier accuracy is 99.4%.

DATASET AVAILABLE AT: http://groupware.les.inf.puc-rio.br/har

Human activities recognition has become a groundwork area of great interest because it has many significant and futuristic applications; including automated surveillance, Automated Vehicles, language interpretation and human computer interfaces (HCI). In recent time an exhaustive and in depth research has been done and progress has been made in this area. The idea of the proposed system is a system which can be used for surveillance and monitoring applications. This paper presents a part of newer Human activity/interaction recognition onto human skeletal poses for video surveillance using one stationary camera for the recorded video data set. The traditional surveillance cameras system requires humans to monitor the surveillance cameras for 24*7 which is oddly inefficient and expensive. Therefore, this research paper will provide the mandatory motivation for recognizing human action effectively in real-time (future work). This paper focuses on recognition of simple activity like walk, run, sit, stand by using image processing techniques.

Rapid growth in numbers of connected devices including sensors, mobile, wearable, and other Internet of Things (IoT) devices, is creating an explosion of data that are moving across the network. To carry out machine learning (ML), IoT data are typically transferred to the cloud or another centralized system for storage and processing; however, this causes latencies and increases network traffic. Edge computing has the potential to remedy those issues by moving computation closer to the network edge and data sources. On the other hand, edge computing is limited in terms of computational power and thus is not well suited for ML tasks. Consequently, this paper aims to combine edge and cloud computing for IoT data analytics by taking advantage of edge nodes to reduce data transfer. In order to process data close to the source, sensors are grouped according to locations, and feature learning is performed on the close by edge node. For comparison reasons, similarity-based processing is also considered. Feature learning is carried out with deep learning: the encoder part of the trained autoencoder is placed on the edge and the decoder part is placed on the cloud. The evaluation was performed on the task of human activity recognition from sensor data. The results show that when sliding windows are used in the preparation step, data can be reduced on the edge up to 80% without significant loss in accuracy.

Rapid growth in numbers of connected devices including sensors, mobile, wearable, and other Internet of Things (IoT) devices, is creating an explosion of data that are moving across the network. To carry out machine learning (ML), IoT data are typically transferred to the cloud or another centralized system for storage and processing; however, this causes latencies and increases network traffic. Edge computing has the potential to remedy those issues by moving computation closer to the network edge and data sources. On the other hand, edge computing is limited in terms of computational power and thus is not well suited for ML tasks. Consequently, this paper aims to combine edge and cloud computing for IoT data analytics by taking advantage of edge nodes to reduce data transfer. In order to process data close to the source, sensors are grouped according to locations, and feature learning is performed on the close by edge node. For comparison reasons, similarity-based processing is al...

In recent years, there has been a proliferation of works on human action classification from depth sequences. These works generally present methods and/or feature representations for the classification of actions from sequences of 3D locations of human body joints and/or other sources of data, such as depth maps and RGB videos.

This survey highlights motivations and challenges of this very recent research area by presenting technologies and approaches for 3D skeleton-based action classification. The work focuses on aspects such as data pre-processing, publicly available benchmarks and commonly used accuracy measurements. Furthermore, this survey introduces a categorization of the most recent works in 3D skeleton-based action classification according to the adopted feature representation.

This paper aims at being a starting point for practitioners who wish to approach the study of 3D action classification and gather insights on the main challenges to solve in this emerging field.

A number of review or survey articles have previously appeared on human action recognition where either vision sensors or inertial sensors are used individually. Considering that each sensor modality has its own limitations, in a number of previously published papers, it has been shown that the fusion of vision and inertial sensor data improves the accuracy of recognition. This survey article provides an overview of the recent investigations where both vision and inertial sensors are used together and simultaneously to perform human action recognition more effectively. The thrust of this survey is on the utilization of depth cameras and inertial sensors as these two types of sensors are cost-effective, commercially available, and more significantly they both provide 3D human action data. An overview of the components necessary to achieve fusion of data from depth and inertial sensors is provided. In addition, a review of the publicly available datasets that include depth and inertial data which are simultaneously captured via depth and inertial sensors is presented.

Human action recognition has a wide range of applications in-cluding biometrics, surveillance, and human computer interaction. The use of multimodal sensors for human action recognition is steadily increasing. However, there are limited publicly available datasets where depth camera and inertial sensor data are captured at the same time. This paper describes a freely available dataset, named UTD-MHAD, which consists of four temporally synchronized data modalities. These modalities include RGB videos, depth videos, skeleton positions, and inertial signals from a Kinect camera and a wearable inertial sensor for a comprehensive set of 27 human actions. Experimental results are provided to show how this database can be used to study fusion approaches that involve using both depth camera data and inertial sensor data. This public domain dataset is of benefit to multimodality research activities being conducted for human action recognition by various research groups. The dataset can be downloaded from http://www.utdallas.edu/~kehtar/UTD-MHAD.html

The emergence of intelligent devices equipped with sensors and connectivity modules has led to the introduction of the smart applications era. Healthcare is one of the major beneficiary of this technology, particularly through mobile health applications that suggest a distributed model for remote monitoring of individuals’ health states and activities, for instance patients, also called remote patient
monitoring. Human activity recognition systems are a set of computing tools that aim to identify the daily activities of persons in order to enhance the quality of their life. They can eventually support medical decisions which is the objective of this project.

Apart from studying the various computational approaches and technics involved in both the remote monitoring and the activity recognition processes, we propose a mobile solution aimed to medical follow-up. It will be designed to work with connected sensors and use distributed machine learning to optimize the smartphones usage. This implies some privacy breaches that we will address through an adequate privacy-preserving protocol. Our solution allows to recognize 12 ambulatory activities with an overall accuracy of 86%. It also has the ability to detect two types of arrythmias.

Video security is becoming more and more vital today due to rapid development of hardware equipments as the number of installed cameras can confirm. This paper presents the system architecture of IVSS, an Intelligent Video Surveillance System based on IPcameras and deployed in an academic environment. In fact, Video surveillance is increasingly found in academic institutions. It is used to oversee the safety of teachers and students, as well as to protect assets from vandalism and theft. In this work the surveillance system is deployed in our university environment, it is based on a set of digital IP video cameras linked by the university IP network infrastructure. Our system architecture is based on efficient moving object detecting and tracking algorithm and a robust statistical activity recognition framework based on SVM which is used for modeling activities. The experimental results on real-time video streams show the feasibility of our system and its effectiveness in human acti...

The recent advancement of deep learning methods has seen a significant increase in recognition accuracy in many important applications such as human activity recognition. However, deep learning methods require a vast amount of sensor data to automatically extract the most salient features for activity classification. Therefore, in this paper, a unified generative model is proposed to generate verisimilar data of different activities for activity recognition. The proposed generative model not only able to generate data that have a similar pattern, but also data with diverse characteristics. This allows for data augmentation in activity classification to improve the overall recognition accuracy. Three similarity measures are proposed to assess the quality of the synthetic data in addition to two visual evaluation methods. The proposed generative model was evaluated on a public dataset. The training data was prepared by systematically varying the combination of original and synthetic data. Results have shown that classification using the hybrid training data achieved a comparable recognition accuracy with the classification using the original training data. The performance of the classifiers maintained at the recognition accuracy of 85%. ABSTRACT The recent advancement of deep learning methods has seen a significant increase in recognition accuracy in many important applications such as human activity recognition. However, deep learning methods require a vast amount of sensor data to automatically extract the most salient features for activity classification. Therefore, in this paper, a unified generative model is proposed to generate verisimilar data of different activities for activity recognition. The proposed generative model not only able to generate data that have a similar pattern, but also data with diverse characteristics. This allows for data augmentation in activity classification to improve the overall recognition accuracy. Three similarity measures are proposed to assess the quality of the synthetic data in addition to two visual evaluation methods. The proposed generative model was evaluated on a public dataset. The training data was prepared by systematically varying the combination of original and synthetic data. Results have shown that classification using the hybrid training data achieved a comparable recognition accuracy with the classification using the original training data. The performance of the classifiers maintained at the recognition accuracy of 85%.

Unobtrusive human activity monitoring using cheap and widely available sensors are the future for human activity recognition. It will support the extensive penetration of new applications in Ambient Assisted Living (AAL), Smart Homes (SH), Smart Cities (SC) and Health Monitoring (HM). The biggest challenges in these applications are the automatic processing and analyzing the large amounts of sensory data as well as building machine learning models for monitoring, detection, recognition and prediction of an activity, movement, state or an event. The aim of this research is to develop a data collection system that will enable detection and monitoring of human activity using very low-cost, unobtrusive passive infrared and microwave radar sensors. Our data collection module is composed of Arduino microcontroller, SD card module and real time clock module and enables connecting several sensors which measurements are to be logged. In our experiments we used a modified microwave radar sensor RCWL-0516 and a modified passive infrared sensor HC-SR501. Both are extremely low cost, easily accessible sensors usually used for general purpose applications like motion detection for light switching. Both sensors were modified in a way to make the raw analog output of the sensor available for logging by the microcontroller. The data collection module enables collecting measurements of up to 4 analog (10-bit precision), and up to 8 digital sensor inputs with sampling rates of up to 200 samples per second. The measurements are logged on a SD card including a precise timestamp that will enable the logs of several modules to be joined together keeping the time alignment of the readings. A separate setup for synchronized initialization of the RTC modules of the separate sensor modules is also presented. A series of experiments in a control environment with volunteers were conducted and the collected data from the sensors are pre-processed and labelled for further analysis and application of machine learning based approaches for automatic recognition and monitoring of human activity.

The contamination of groundwater by heavy metal, originating either from natural soil sources or from anthropogenic sources is a matter of utmost concern to the public health. Remediation of contaminated groundwater is of highest priority since billions of people all over the world use it for drinking purpose. In this paper, thirty five approaches for groundwater treatment have been reviewed and classified under three large categories viz chemical, biochemical/biological/biosorption and physico-chemical treatment processes. Comparison tables have been provided at the end of each process for a better understanding of each category. Selection of a suitable technology for contamination remediation at a particular site is one of the most challenging job due to extremely complex soil chemistry and aquifer characteristics and no thumb-rule can be suggested regarding this issue. In the past decade, iron based technologies, microbial remediation, biological sulphate reduction and various adsorbents played versatile and efficient remediation roles. Keeping the sustainability issues and environmental ethics in mind, the technologies encompassing natural chemistry, bioremediation and biosorption are recommended to be adopted in appropriate cases. In many places, two or more techniques can work synergistically for better results. Processes such as chelate extraction and chemical soil washings are advisable only for recovery of valuable metals in highly contaminated industrial sites depending on economical feasibility.► A number of groundwater treatment technologies for heavy metal remediation have been reviewed. ► Iron based techniques, Biological Sulphate Reduction and Permeable Reactive Barriers were most successful in the past decade. ► Biosorption is the emerging field mostly experimented in laboratories. ► More than one techniques can be coupled for site specific problems for better performance.

Portable navigation is increasingly becoming an essential part of our lives. Knowing a person’s position and velocity using a portable device, such as a smartphone, tablet, smartwatch, or smartglasses, has become a basic part for many people’s lives. However, portable navigation is still facing many challenges: mainly navigation in urban canyons or indoors where Global Navigation Satellite System (GNSS) signals are weak or unavailable, and the need to compromise the accuracy and precision of inertial sensors in return for limiting their cost and size to place in portable consumer devices. As a result, portable navigation needs to employ different algorithms for each mode of motion, to assure optimal results in each case. This thesis’ contribution is to provide a motion mode recognition module that aims to detect a wide range of motion modes using micro-electro-mechanical sensors (MEMS) within a portable navigation device, which is robust to the device usage, and device orientation. The proposed motion mode recognition module works with good accuracy in the absence of absolute navigational signals (such as GNSS or WiFi). The motion modes detected are: stationary, walking, running, cycling, in a land-based vessel, walking in a land-based vessel, standing on a moving walkway, walking on a moving walkway, moving on stairs, taking elevator, standing on an escalator, and walking on an escalator. The motion mode recognition module involves the following steps: data input reading, pre-processing, feature extraction, classification, and post-classification refining techniques. During the data input reading step, signal readings from a group of sensors, are simply obtained. In the second step, pre-processing, such raw signals are processed and fused to obtain more meaningful variables. Feature extraction involves grouping the variables into a window, and extracting a group of features from that window. Classification is the final step where the feature vector is fed to predict the motion mode. To train and evaluate the classifier models, more than 2400 tests were conducted using various portable navigation devices, such as smartphones, tablets, smartwatches, and head mounted systems. The training tests involved more than 35 users of various genders, weights, heights, and ages, and covering various device usages and orientations.

Human Motion Analysis (HMA) is currently one of the most popularly active research domains as such significant research interests are motivated by a number of real world applications such as video surveillance, sports analysis, healthcare monitoring and so on. However, most of these real world applications face high levels of uncertainties that can affect the operations of such applications. Hence, the fuzzy set theory has been applied and showed great success in the recent past. In this paper, we aim at reviewing the fuzzy set oriented approaches for HMA, individuating how the fuzzy set may improve the HMA, envisaging and delineating the future perspectives. To the best of our knowledge, there is not found a single survey in the current literature that has discussed and reviewed fuzzy approaches towards the HMA. For ease of understanding, we conceptually classify the human motion into three broad levels: Low-Level (LoL), Mid-Level (MiL), and High-Level (HiL) HMA.

The technology of Smart Homes (SH), as an instance of ambient assisted living technologies, is designed to assist the homes’ residents accomplishing their daily-living activities and thus having a better quality of life while preserving their privacy. A SH system is usually equipped with a collection of inter-related software and hardware components to monitor the living space by capturing the behaviour of the resident and understanding his activities. By doing so the system can inform about risky situations and take actions on behalf of the resident to his satisfaction. The present survey will address technologies and analysis methods and bring examples of the state of the art research studies in order to provide background for the research community. In particular, the survey will expose infrastructure technologies such as sensors and communication platforms along with artificial intelligence techniques used for modeling and recognizing activities. A brief overview of approaches used to develop Human–Computer interfaces for SH systems is given. The survey also highlights the challenges and research trends in this area.

Human Activity Recognition (HAR) has attracted much attention from researchers in the recent past. The intensification of research into HAR lies in the motive to understand human behaviour and inherently anticipate human intentions. Human activity data obtained via wearable sensors like gyroscope and accelerometer is in the form of time series data, as each reading has a timestamp associated with it. For HAR, it is important to extract the relevant temporal features from raw sensor data. Most of the approaches for HAR involves a good amount of feature engineering and data pre-processing, which in turn requires domain expertise. Such approaches are time-consuming and are application-specific. In this work, a Deep Neural Network based model, which uses Convolutional Neural Network, and Gated Recurrent Unit is proposed as an end-to-end model performing automatic feature extraction and classification of the activities as well. The experiments in this work were carried out using the raw data obtained from wearable sensors with nominal pre-processing and don’t involve any handcrafted feature extraction techniques. The accuracies obtained on UCI-HAR, WISDM, and PAMAP2 datasets are 96.20%, 97.21%, and 95.27% respectively. The results of the experiments establish that the proposed model achieved superior classification performance than other similar architectures.

The release of smart phones equipped with a rich set of sensors has
enabled human activity recognition on mobile platforms. Monitoring the daily
activities and their levels helps in recognizing the health and wellness of the
users as an ideal application. Ubiquity of the mobile phones, unobtrusiveness,
ease of use, communication channels and playfulness make mobile phones a
perfect platform also for inducing behavior change for a healthier and more
active lifesytle. In this paper, we provide a review on the activity recognition
systems that use integrated sensors in the mobile phones with a special focus on
the systems that target personal health and wellbeing applications. Initially, we
provide background information about the activity recognition process, such
as the sensors used, activities targeted, the steps of activity recognition using
machine learning algorithms before listing the challenges of activity recognition
on mobile phones. Next, we focus on the classification of existing work on the
topic together with a detailed taxonomy. Finally, we investigate the directions
for future research.

Human activity monitoring technologies are one of the essential systems for elderly care. Advances in electronic systems, sensor technologies, and communication network protocols have enabled a new generation of integrated health systems to be created. The solution presented in this document represents an integrated system prototype that provides an efficient technological tool to caregivers operating promptly and ensures efficient performance throughout the entire healthcare system process. This solution differs from previous works for the coexistence of a series of innovative aspects. Human activity recognition is based on combining different types of information: environmental data, physiological data, inertial data, and indoor location data of patients; CNN network for locating position and activities; Virtual Reality System (VR) for optimizing the neural network and related training.

In this paper we propose the use of a widely known paradigm in the forensic area to detect and categorize human activities. 4W1H describes the context in any environment by describing it using 5 base variables, Who, When, What, Where and How. We make use of this description plus an intention variable known as “Why” to be able to predict and react accordingly to the current user’s situation. We show the hardware setting required to detect these variables as well as some approaches to sense them from the environment using a minimal hardware setting. We also look into the current work in categorizing the data using Self Organizing Maps and Clustering Techniques.

The aim of this paper is to present a novel indoor human activity recognition system designated for ambient assisted living environments. To overcome the drawbacks of current solutions in terms of equipment costs and computational complexity, we propose a simple yet robust activity recognition system that combines the output of an activity classification algorithm using smartphone sensors such as accelerometer and gyroscope with a Beacon-based indoor localization predictor to detect more complex activities based on a decision rule system. For the multiclass classification algorithm we employ a ConvL-STM algorithm, while for the positioning system we propose an ensemble based solution combining Multilayer Perceptron with Gradient Boosted Regression and k Nearest Neighbors. Our solution has been tested in a controlled home environment setup achieving an average localization error of 0.4m while for the final activity recognition system we report an accuracy of 91.0% concluding that the proposed solution is accurate enough to be integrated as a monitoring tool facility within ambient assisted living systems.

Activity detection is becoming an integral part of many mobile applications. Therefore, the algorithms for this purpose should be lightweight to operate on mobile or other wearable device, but accurate at the same time. In this paper, we develop a new lightweight algorithm for activity detection based on Long Short Term Memory networks, which is able to learn features from raw accelerometer data, completely bypassing the process of generating hand-crafted features. We evaluate our algorithm on data collected in controlled setting, as well as on data collected under field conditions, and we show that our algorithm is robust and performs almost equally good for both scenarios, while outperforming other approaches from the literature.

When dealing with an elevated sensor number we may
have to reduce our sampling time or the sensing features in order to achieve real time sensing. In this paper we propose a make use of the recently developed theory of Compressive Sensing to try and reduce the number of sensing as well as the sample size without loosing quality in our sampling for doing a good feature recognition.

Recognition of the mode of motion or mode of transit of the user or platform carrying a device is needed in portable navigation, as well as other technological domains. An extensive survey on motion mode recognition approaches is provided in this survey paper. The survey compares and describes motion mode recognition approaches from different viewpoints: usability and convenience, types of devices in terms of setup mounting and data acquisition, various types of sensors used, signal processing methods employed, features extracted, and classification techniques. This paper ends with a quantitative comparison of the performance of motion mode recognition modules developed by researchers in different domains.

To show the impact of human activity on biodiversity in humid equatorial region, termites, insects recognized as bio-indicators of climate change in the tropics were studied in two habitats: The National Center for Floristic (CNF), a preserved habitat for 51 years and the Campus of Cocody (CC), a strongly anthropic environment. Termites were sampled using a standardized method designed for rapid assessment of termite. In both areas, 18 species were collected, 17 species in CNF and 7 species in the Campus area. Significant change in diversity was found between the CNF and the Campus area. In addition, the relative abundance of termites showed a significantly greater decline in the Campus area. The highest abundance of termites was obtained in the CNF (2.95 ind./sections) unlike the Campus area where the relative abundance was 1.55 ind./section. Shannon and Simpson diversity index recorded in the CNF was higher than those of the Campus area. The low similarity index calculated (0.38) showed that these two areas were strongly different. The soil feeders seem to be the most sensitive group to human activities. The CNF preserved of human activities allowed reconstituting different species of termite colonies, unlike the Campus area which still subject to strong anthropogenic activities.

In this paper, we studied Stacked Denoising Autoencoder(SDA) model for Human pose-based action recognition. We used public dataset Chalearn 2013 which contains Italian body language actions from 27 persons. We studied two model of SDA for pose clustering: 1) Traditional SDA with epoch and Neural Network supervised classifier and 2) Marginalized SDA which faster and ELM supervised classifier. We used supervised classifier by using initial cluster data from K-means. We deployed global tuning that updating the weight during iterative learning.