Smartphone Supported Activity Level Estimation

This thesis deals with the prototypical implementation of activity zone monitoring using the mobile phone. It uses the mobile phone sensors especially the accelerometer sensor to establish four types of motions and through a corresponding analysis with heart rate monitoring equipment, and then establishes the intensity of the activity. The goal behind the activity zone estimator as being suggested in this thesis is to provide notions of an activity of a certain intensity by only using the mobile phone without using external sensors. The implementation is based on first an analysis of existing technologies both when it comes to programming and when it comes to applications being available for mobile phones, and our implementation analysis then points out that the accelerometer is well tailored to establish an activity zone. However, the challenges might still occur with respect to, for example, the position of the smartphone on user's body, elevation of the ground where user performs the activity, and battery life time. The objective of this thesis is to establish a prototypical implementation of mobile sensors to collect activity information from the users. This is done by creating the state of the art overview of mobile supported activity monitoring technologies. Our prototypical implementation is called zonEstimator (ZE) that classifies acceleration data into four states of motion; slow walking, fast walking, slow running and fast running, and through the states of motion indicates the activity zone and the activity intensity. To analyse the correlation between heart rate values retrieved by an external device (Zephyr sensor device) and acceleration results collected by the user's mobile phone sensors. To examine how much we can get about a user's activity zone by benefiting from only built-in accelerometer sensor of a smart-phone without using an external sensor in his daily life. The components used in this thesis are a smart phone with the internal accelerometer sensor, both used for data collection and analysis, and an external heart rate sensor being used for having the de facto measurement of the heart rate intensity, and thus being able to correlate the mobile phone reported values with the measured heart rate values. The developed application consists of three main modes; training, application, and reporting. The training mode includes both the heart rate sensor and internal accelerometer sensor, and this is the first step to create the aforementioned correlation between acceleration and heart rate measurements. The application mode hosts only the built in sensor of the smart phone, which is to provide the real time results of activity zone of the user, and the reporting mode is to provide history of the results for the user. The accuracy of using built-in sensors for detecting the activity zone of a user changes under some specific circumstances. These relate to the position of the smart-phone on user's body, the slope of the ground where the activity is performed, physical condition of the user and some other factors. The accuracy of the detection is about 90\% when the user carries the phone by his palm while it is greater than about 75\% when it is carried by an arm band for fast walking and slow running activity types. We have found out that average acceleration goes up about 0.14 for each heart beat, meaning that as the intensity of the activity increases acceleration increases as well for the aforementioned activity types. We have observed a strong correlation between the acceleration of the smart phone and heart rate measurements of the user when performing the following activities; slow walking, fast walking, slow running and fast running. The accuracy of the correlation is strongly associated with the position of the phone, and the incline of the ground where the user performs the activity. Plus, training phase is vital for accurate conclusions since the results vary from user to user in terms of age, sex, height, weigh and from smart phone to smart phone. Our analysis thus shows that the smart phone supported activity monitoring needs a training set per user. Potential way ahead would be to use mobile phone sensors to establish a better judgement of the activity and then fine grained analysis to establish the zone estimator. Our recommendation is that multi sensor analysis fits well in order to find out what the user is doing whether walking or running. However, it might be heavy for the mobile phone to measure accelerometer sensor data all the time. Therefore, there might be the idea of using an external accelerometer sensor such as the accelerometer that the Fitbit device uses, that could be used for getting a better idea of intensity of the activities. However, we may have the problem of Bluetooth this time, which might be an interesting discussion which one may look into