You are here
Accelerometer-based gait analysis using mobile phones
There are a lot of various sensors in modern smartphones, such as an ambient light sensor, a proximity sensor, Global Positioning System (GPS), a compass, a gyroscopic sensor, an accelerometer and others. In this work we study potentialities of accelerometers, specifically gait analysis methods. Some of them have already implemented and been used successfully, such as pedometer. Other potentialities, in which we interested, are human identification and health monitoring. A number of biometric methods of identification have been introduced over the years, such as eye scans (retina or iris), fingerprint recognition, hand or palm geometry, voice recognition, facial recognition, heartbeat biometrics and others. But most of them either require expensive and specialized hardwares or are inefficient as often inexpensive systems are prone to errors of the second kind (for instance, builtin laptops fingerprint readers). Therefore we've decided to study the ability to identify human using his gait. It is inexpensive because of using built-in MEMS-based accelerometers and also requires no specific actions from the user except just walking. There are many articles on the topic of accelerometer based gait authentication, such as [1, 2, 3]. Other our interest is health monitoring, which now also is very topical and important. The works in this area are [4, 5, 6].
1. Mohammad O. Derawi, Patrick Bours, Kjetil Holien. “Improved Cycle Detection for Accelerometer Based Gait Authentication”, 2010
2. Murat Ekinci “A New Approach for Human Identification Using Gait Recognition”, 2006
3. Mostayed, A.; Mynuddin, M.; Mazumder, G.; Sikyung Kim; Se Jin Park; Dept. of Electr. Eng., Kongju Nat. Univ., Kongju. “Abnormal Gait Detection Using Discrete Fourier Transform”, 2010
4. Yu-Jin Hong, Ig-Jae Kim, Sang Chul Ahn, Hyoung-Gon Kim. “Mobile health monitoring system based on activity recognition using accelerometer”, 2009
5. Joshua Juen, Bruce Schatz. “Health Monitoring using Gait Analysis from Smart Phone Accelerometers”, 2011
6. Alexey Zhilyaev “Biomechanical and electrophysiological criteria of estimation of the functional state of the musculoskeletal system of the lower extremities”, 2003 (In Russian)
Our current ideas how to use accelerometer are to determine pattern of movement and to identify human: 1. Patterns of movement Our purpose is to calculate the rate of movement, but now it's very difficult task, so we use patterns. Movement is divided into three patterns: human isn't walking, human is walking and human is running. To determine pattern of movement we use FFT (Fast Fourier Transform). Depending on pattern we can, for instance, switch phone's modes, such as sound volume. 2. Identification The main idea is that certain identifier, called the Gate Data (GD), is calculated for every walking person. Then the GD of this person is compared with database of other GD and the system makes a decision on further actions. The principal calculations of getting GD are performed in the frequency domain. Therefore, the algorithm is also based on FFT. The algorithm of getting GD is shown in the Figure1. Since FFT calculations are very complex for smartphone's processors we perform calculations on the server: smartphone transmit accelerometer's data to server via WiFi.However, we have some problems with this system that we want to eliminate during this project. This problems connected with both the inaccuracy of device (we work with Nokia N900) and some domestic reasons, for instance, other clothes and footwear can change the gait. The problems with device lie in the quantization of the accelerometer's data that can't be changed. Therefore we should update our algorithm for this case.
Grankin Maxim, SUAI
Khavkina Elizaveta, SUAI
Ometov Alexander, SUAI
3rd quarter 2012:
- Development of algorithms of human identification by his gait with the help of MEMS that built in mobile phones.
- Software development for continuous data transmission from mobile phone to computer via Wi-Fi.
- Integration of the developed algorithm in software.
- Development of algorithm of determination pattern of movement.
4th quarter 2012:
- Development of bundled software for the continuous user identification by his gait.
- Testing of developed bundled software with various categories of probationers.
1st quarter 2013:
- Biomechanical research of diagnostic methods of functional state of the main joints of the lower extremities.
- Adding features of the continuous data transmission from mobile phone to computer via 3G to the software.
2nd quarter 2013:
- Development of diagnostic algorithms of the functional state of the main joints of the lower limbs with the help of MEMS that built in mobile phones.
- Testing and finalization of the developed software.