Human Activity Recognition in Android
Download as PPTX, PDF4 likes10,089 views
This document discusses the development of an Android application for physical activity recognition using the accelerometer sensor. It provides background on the Android operating system and its open development environment. It then summarizes relevant research papers on activity recognition using mobile sensors. The document outlines the process of collecting and labeling accelerometer data from smartphone sensors during different physical activities. Features are extracted from the sensor data and several machine learning classifiers are evaluated for activity recognition. The application will recognize activities and track metrics like calories burned, distance traveled, and implement fall detection and medical reminders.
Human Activity Recognition in Android
- 1. Made by: Surbhi Jain Aishwarya Jain
- 3. The Android Framework The Android platform is an open platform for mobile devices consisting of an operating system, applications and middleware Android gives users the opportunity to build and publish their own applications by providing an open development environment. Android treats all applications (native and third-party) as equals. Therefore, having such an open development environment requires security measures to be taken in order to protect the integrity of the Android platform and the privacy of its users.
- 4. • Android is an open source mobile operating system with Linux kernel. • The Android SDK is installed in to Eclipse. • Android treats both native and third party applications as the same. So we can build and develop our own applications easily here. • The android software development kit includes a set of development tool such as a debugger, libraries, handset emulator, documentation, sample code and tutorials. • Android SDK has a java frame work and a powerful API for the hardware embedded on smartphones. Why Android ?
- 6. Various Android Sensors The Android has several sensors available: • Accelerometer • Orientation • Ambient Light • Proximity • Magnetic Force Use of these sensors does not require direct user permission!
- 8. Activity Recognition Desired Outputs: - Physical Activities (e.g., Running, Walking) - Approximate time spans - Quick detection of change Accelerometer Usage
- 9. Our Objective To explore the Accelerometer as a measure of context- aware based applications for physical activity recognition on the Android framework.
- 10. Literature Survey List Of Paper Studied: • Paper 1 : Applications of Mobile Activity Recognition Authors : Jeffrey W. Lockhart, Tony Pulickal AND Gary M. Weis • Paper 2 : User, Device and Orientation Independent Human Activity Recognition on Mobile Phones: Challenges and a Proposal Authors : Yunus Emre Ustey, Ozlem Durmaz Incel AND Cem Ersoy • Paper 3 : Simple and Complex Activity Recognition Through SmartPhones Authors : Das, B., Krishnan, Narayanan C., Thomas, B.L. AND Cook, D.J • Paper 4 : Fall Detection by Built-In Tri-Accelerometer of Smartphone Authors : Yi He, Ye Li AND Shu-Oi Bao • Paper 5 : Feature Selection Based On Mutual Information For Human Activity Recognition Authors : Khan, A., Chehade, N.H., Chieh Chien AND Pottie. G
- 11. • Paper 6 : Smartphone-based Monitoring System for Activities of Daily Living for Elderly People and Their Relatives Etc. Authors : Kazushige Ouchi AND Miwako Doi • Paper 7 : Environment Feature Extraction and Classification for Context Aware Physical Activity Monitoring Authors : Troped, P.J., Evans,J.J. AND Pour,G.M • Paper 8 : Fall Detection based on movement in Smartphone Technology Authors : Gueesang Lee AND Deokjai Cho • Paper 9 : Activity logging using lightweight classification techniques in mobile devices Authors : Henar Martı´n ,Ana M. Bernardos ,Josue´ Iglesias • Jose´ R.Casar • Paper 10 : Privacy control in smart phones using semantically rich reasoning and context modeling Authors : Dibyajyoti Ghosh, Joshi, A., Finin, T. AND Jagtap, P contd..
- 12. • Paper 11 : Towards Successful Design of Context-Aware Application Frameworks to Develop Mobile Patient Monitoring Systems Using Wireless Sensors Authors : Al-Bashayreh, M.G. Hashim, N.L. AND Khorma, O.T • Paper 12 : ActivityMonitor: Assisted Life Using Mobile Phones Authors : Matti Lyra AND Hamed Ketabdar
- 13. Comparison among the papers Paper Parameters Used Algorithm Used/Proposed Paper 1 Nil Neural networks and J48 decision trees Autocorrelation, K-nearest neighbors (KNN), Paper 2 Mean, Variance, Std. Dev, fast Fourier transform (FFT) coefficients Zero Crossing Rate, Period Mean, min, max, Std. Dev, Multi-layer Perceptron, Paper 3 Zero Crossing Rate, correlation Naïve Bayes, Bayesian network, Decision Table, Best-First Tree, and K-star Acceleration due to Signal Magnitude Vector, Signal Magnitude Paper 4 body movement; 2) Area (SMA), Tilt Angle (T A) gravitational acceleration, median filter Standard deviation, Mean, Tree-based, feature selection algorithm based Paper 5 Absolute mean, Energy ratio, on mutual information, binary Ratio of DC to sidelobe, First decision-tree with a naıve Bayes classifier sidelobe location, Max value, Short time energy, Correlation Paper 6 Average, minimum, maximum Stochastic model, Neural Networks, SVM every and variance, MFCC (Mel- 1 sec. Frequency Cepstral Coefficient), RMS (Root Mean Square) and ZCR (Zero-Crossing Rate)
- 14. Paper 7 Mean and sigma K-nearest neighbor of the Gaussian function Paper 8 Lower NIL Threshold (LT) and Upper Threshold (UT). Paper 9 Mean, Variance, Zero crossing Naıve Bayes, rate ,75 percentile Decision Table and Decision Tree Paper 10 NIL NIL Paper 11 NIL NIL Paper 12 Average magnitude value, Multi-Layer Perceptron (MLP) average rate of change, weighted sum
- 15. Current Problems • First and Foremost is the use of Body Worn sensors. Today, In most of the apps we have found that external sensors are used to detect the physical movements of a person. Practically it is not possible to carry an external device with you Sometimes people forget to wear the device. • In most of the apps Positioning of the device is the concerned for the success of application i.e. Most of the apps build are position specific of device. If the device is kept in hands then values will be different from the values generated when the device is kept in pocket. • Use of multiple Sensors to achieve the same goal which makes the application bulky leading to slower processing of the data and also affects its cost.
- 16. Restating the Problem We primarily focused on the Activity Recognition project Inputs: -X acceleration -Y acceleration -Z acceleration Desired Outputs: - Physical Activities (e.g., Running, Walking) - Approximate time spans - Quick detection of change
- 17. The Activity Recognition Process
- 18. Data Collection Process The first step to the project was to collect raw accelerometer data and transform it into features that WEKA, the machine-learning tool that we implemented, used to train a classifier. To accomplish this, we first took in sensor samples made up of acceleration readings in the x, y, and z directions and computed their magnitudes. Labeled all the data manually in terms of running, walking, standing, sitting. To make the data more accurate data of more than 20 minutes have been taken. Data gathering was done by performing experiments on four subjects. Each of the four subjects were asked to collect the data activity one by one by placing smartphone at the positions mentioned above. Each subject performed the set of 6 activities one by one for the duration of two minutes and the respective data was recorded in a .csv file in the external storage of the smartphone. Contd...
- 20. Feature Extraction Feature Extraction is the process of extracting key “features” from a signal. Features will be extracted from every sample window of 512 samples. The following features we Will be using in our project: 1. The Fundamental Frequencies: The average of the three dominant frequencies of the signal over the sample window. This was found via a Discrete Fourier Transformation. 2. Average Acceleration: The arithmetic mean of the acceleration magnitudes over the sample window. 3. Max Amplitude: The maximum acceleration value of the signal in the sample window. 4. Min Amplitude: The minimum acceleration value of the signal in the sample window.
- 21. Classification is the process of labeling unknown patterns based on the knowledge of known patterns of data. Four different classifiers were used: K-Nearest Neighbor: Based on the shortest euclidean distance between the unknown and known data’s feature vector Naïve Bayes: Assumes the absence of one feature does not disqualify a candidate (e.g., an object which is red and round is an apple, even if is not known to be a fruit) J48(Decision Tree): J48 builds decision trees from a set of labeled training data using the concept of information entropy. It uses the fact that each attribute of the data can be used to make a decision by splitting the data into smaller subsets. Random Forest: An ensemble classifier using many decision tree models. It can be used for classification or regression. Classification
- 22. Use Case Diagram
- 23. The application will be divided in several modules which will be implemented time to time. Module-1 Activity Recognition Physical Activites like cycling, running,walking, standing etc performed by the user will be recognized in this module. User will click on the app icon or start button in the app which make him able to run sensors and thus his motion wil be detected. Module-2 Location Based Activity Recognition In addition with Activity Recognition GPS sensor will be used to find the location of user also what activity he is performing at that location. This will help in Fall Detection Module discussed later. Overall Description Contd…
- 24. Module-3 Fall Recognition In addition with physical Movement Recognition, Fall Recognition will be there.. Whenever Fall detection will have positive result then alarm will be raised instatntly and then app will monitor physical activities and if motion is not detected it will send an emergency message to the guardian informing about this accident. Module-4 Physical Activity Chart The User will be able to see his/her daily physical activity chart i.e. how much he has done workouts today and what type of physical activity he/she performed during the day. Module-5 Calorie Burnt The user will have the ability to see the calories burnt by him within a day and how much calories he/she should be burnt to be physically fit. Contd…
- 25. Module-6 Medical Reminder The user can set the medical reminder if he wants. For this he/she have to feed the prescription in the phone with the timings and the made reminder active. Module-7 Distance Travelled The user have the ability to see the distance he travelled by running or by cycling or by walking
- 28. IMPLEMENTATION
- 30. Risk and Mitigation Risk Description Risk Area Prob Impa RE (P* Risk Mitigati Id of Risk ( Identify abilit c I) Select on Plan Risk y t (I) ed if 8 is ‘Y’ Areas for (P) for your Mitiga project) ti on (Y/N) Position of mobile Sensor 1. i.e. Whether the readings High High High Taking data by phone is taken in variation Y considering all the hand or kept in possible locations. chest pocket or in pant’s pocket 2. Battery drainage Hardware High Medi Medium N NIL um 3. Sending each sms Security Medi Low Medium N NIL to a hidden 3rd um Party address Device 4. Computational limitations Hardware Medi Low Medium Y On cloud storage service. um
- 31. Sources International Journal of Distributed Sensor Networks provided by Hindawi.com IEEE Sensors Journal provided by http://www.ieee- sensors.org/journals. IJCA Proceedings on International Conference on Recent Trends in Information Technology and Computer Science 2012 Acoustics, Speech and Signal Processing (ICASSP), 2012 IEEE International Conference Sensors Applications Symposium (SAS), 2013 IEEE 2012 IEEE RIVF International Conference IEEE Symposium on Security and Privacy Workshops@2012 ACM Transactions on Knowledge Discovery from Data (TKDD) ACM Journal of Data and Information Quality
- 32. The End