Project

Human Activity Recognition using Mobile Sensors

Updates
0 new
0
Recommendations
0 new
0
Followers
0 new
0
Reads
1 new
55

Project log

Kishor Walse
added 2 research items
Mobile Phone used not to be matter of luxury only, it has become a significant need for rapidly evolving fast track world. This paper proposes a spatial context recognition system in which certain types of human physical activities using accelerometer and gyroscope data generated by a mobile device focuses on reducing processing time. The benchmark Human Activity Recognition dataset is considered for this work is acquired from UCI Machine Learning Repository, which is available in public domain. Our experiment shows that Principal Component Analysis used for dimensionality reduction brings 70 principal components from 561 features of raw data while maintaining the most discriminative information. Multi Layer Perceptron Classifier was tested on principal components. We found that the Multi Layer Perceptron reaches an overall accuracy of 96.17 % with 70 principal components compared to 98.11 % with 561 features reducing time taken to build a model from 658.53 s to 128.00 s.
In this paper, we have introduced the concept of user modeling which simply means “Doing right thing at right place at right time”. In this paper we discuss firstly about the history of modeling user and hierarchy of user model which will help in designing the user model for different context. Also give the comparison between the stereotype and feature-based user model. By reading this paper one can select the proper user model for the corresponding context aware application. In future the benefit of this modeling is that the mobile will be smart enough that it will handle all the thing that are now handle by the mobile owner.
Kishor Walse
added 2 research items
Nowadays, all smartphones are equipped with powerful multiple built-in sensors. People are carrying these “sensors” nearly all the time from morning to night before sleep as they carry the smartphone all the time. These smartphone allow the data to be collected through built-in sensors, especially the accelerometer and gyroscope give us several obvious advantages in the human activity recognition research as it allow the data to be collected anywhere and anytime. In this paper, we make use of publicly available dataset online and try to improve the classification accuracy by choosing the proper learning algorithm. The benchmark dataset considered for this work is acquired from the UCI Machine Learning Repository which is available in public domain. Our experiment indicates that combining AdaBoost.M1 algorithm with Random Forest, J.48 and Naive Bayes contributes to discriminating several common human activities improving the performance of Classifier. We found that using Adaboost.M1 with Random Forest, J.48 and Naive Bayes improves the overall accuracy. Particularly, Naive Bayes improves overall accuracy of 90.95 % with Adaboost.M1 from 79.89 % with simple Naive Bayes.
Kishor Walse
added 8 research items
Human activity recognition is an extensive area of a machine learning research because of its applications in healthcare, smart environments, homeland security, entertainment, etc. Study for human activity recognition observes that researchers are interested mostly in the daily activities of the human. Activity recognition using sensor data plays an essential role in many applications. Earlier, mostly wearable sensors were used to recognize various daily living activities. But, wearable sensors and environmental sensors both are bulky and costly, so switching towards smartphone sensors seems reliable and easier option to researchers and hence smartphone sensors are now-a-days widely used by researchers. In this paper, we review the studies done that implement activity recognition systems on smartphone using various sensors. We also discuss various facets of these studies.
Human activity recognition is the most recently introduced and nowadays widely used term. Human activity recognition is a big research point. Many researchers had developed different architectural systems for this. In this paper a survey of frameworks designed to develop a human activity recognition technique is done. This paper makes survey of different types of frameworks such as CenceMe, EmotionSence, Cyberguide etc. And the general terms related to the human activity recognition are explained.
Human activity recognition is intrinsic area of exploration just because of its real world's applications. The sensors included smart phones are used to recognize activity. Mobile phone provides small size, CPU, Memory and Battery. A detailed survey of Design classifier for human activity recognition systems by using soft computing techniques are discussed in this report. Also various classifiers are discussing in this survey. Describe supervised and unsupervised learning. Also describe various types of classifiers in this paper.
Kishor Walse
added a research item
Nowadays, all smartphones are equipped with powerful multiple built-in sensors. People are carrying these "sensors" nearly all the time from morning to night before sleep as they carry the smartphone all the time. These smartphone allow the data to be collected through built-in sensors, especially the accelerometer and gyroscope give us several obvious advantages in the human activity recognition research as it allow the data to be collected anywhere and anytime. In this paper, we make use of publicly available dataset online and try to improve the classification accuracy by choosing the proper learning algorithm. The benchmark dataset considered for this work is acquired from the UCI Machine Learning Repository which is available in public domain. Our experiment indicates that combining AdaBoost.M1 algorithm with Random Forest, J.48 and Naive Bayes contributes to discriminating several common human activities improving the performance of Classifier. We found that using Adaboost.M1 with Random Forest, J.48 and Naive Bayes improves the overall accuracy. Particularly, Naive Bayes improves overall accuracy of 90.95 % with Adaboost.M1 from 79.89 % with simple Naive Bayes.