Gait recognition is the next potential biometric which is being looked at very closely as a viable technology of the future. Gait recognition, simply put, is the potential verification and/or identification of an individual based upon their unique stride, or the unique way in which one walks in normal, everyday life. It can be considered both a behavioral and physical based potential biometric, and thus is considered part of the field known as kinematics, which is the study of motion.
With our gait, the various aspects of the body closely position themselves at unique lengths, angles, and speeds. At the present time, there is strong interest in using gait recognition for identification purposes, because the raw data needed to create the biometric templates can be collected in a very covert fashion. As a result, gait recognition can be used very well with CCTV technology and facial recognition.
The process behind gait recognition
There are currently two methods which are being investigated for capturing and examining the unique strides in individuals:
- Static shapes: This methodology actually takes a silhouette pattern of a walking person from a series of vide of video frames taken by a CCTV camera. This image of the walking individual is defined in a boxed region of the video frame, and any extraneous features from the external environment are removed. The silhouette image is then created, and from there, a further series of images are created, then digitally mapped to yield what is known as the gait signature;
- Dynamic shapes: This type of methodology uses an accelerometer, placed on the individual’s leg, and the stride (walking) motion is then collected and analyzed from three different directions.
There are numerous methods which are currently being investigated to capture the raw data. They are as follows:
- Machine vision: With this, the individual video frames which display the individual walking are isolated from one another frame by frame.
- Doppler radar: With this system, a sound wave is bounced off of the leg of the individual, and the return sound wave is then measured. This is done by ascertaining the velocity of the signal which is modulated by the moving parts of the leg.
- Thermography: This technique is embedded into a computer vision system in order to cast out the moving leg in difficult and extraneous lighting situations.
After the raw images are captured and collected, the next step in the biometric process is unique feature extraction. Currently, there are two methodologies which are also being examined to accomplish this task:
- Model based analysis: With this technique, the height of the individual, the distance between the head and the pelvis, the total distance between the pelvis and the feet, as well as the total distances between the moving feet are measured and calculated. Also, the relationships of the distances and the angles of the walking individual can be converted into various vectors as well.
- Holistic analysis: With this technique, the silhouette of the walking individual is also calculated and measured.
Finally, after unique feature extraction, the next step is the creation of the biometric templates. In this regard, gait recognition is different from the other biometric technologies in that actual sequences of movements are compared against one another. There are also various template creation techniques which are under research and development, and they are as follows:
- Direct matching: With this approach, a gait sequence is identified by comparing the raw data which is captured with the reference data, and from there, the various distances which yield the least minimal distances are subsequently calculated.
- Dynamic time warping: With this technique, the closeness between two sequences of gait movements is measured. The variables examined and analyzed include the time, speed, and the various types and kinds of walking patterns.
- Hidden Markov models: This is a statistical based methodology in which the probability of the similarity of the shapes which appear in the walking stride of an individual are examined and analyzed in succession.
Gait Recognition: The advantages & disadvantages
Gait recognition, although still undergoing research and development, can be compared and evaluated against the same seven criterion the other biometric technologies have been rated against:
- Universality: The collection of raw data requires that an individual possess a walking stride which is basically unimpeded and be without the use of such walking aids such as crutches, walkers, and even wheelchairs;
- Uniqueness: Since gait recognition is not yet a viable biometric technology, it’s true, absolute uniqueness still cannot yet be ascertained when compared to the other biometric technologies, and as a result, at the present time, gait recognition is predicted to work very well in a potential multimodal biometric security solution;
- Permanence: A person’s walking stride can be affected by such variables as older age, physical injuries, or even disease. The walking stride can also be affected by such variables as the type of surface the individual is walking upon, any additional or extraneous weight which is being carried by the individual, and even the type and kind of footwear the individual uses while they are walking;
- Collectability: A potential advantage of gait recognition is that raw data which has a low resolution as captured by the CCTV system can still be used to create robust biometric templates;
- Performance: Some of the studies conducted thus far have claimed that gait recognition possesses an Equal Error Rate (ERR) of 5% (and perhaps even less);
- Acceptability: It appears that gait recognition can be considered to be potentially a noninvasive biometric, but the issues of privacy rights and civil liberties violations could emerge as a person’s unique walking stride can be collected and analyzed covertly;
- Resistance to circumvention: At this point in research and development, it is deemed that it could be very difficult to mimic another individual’s walking stride.
Finally, it is forecasted that gait recognition can be used as a great surveillance tool (especially when combined with CCTV technology and facial recognition) in large venue settings, such as sporting events, concerts, transportation hubs, educational institutions, etc. Also, gait recognition can be used to track suspicious behaviors with regards to motion of an individual, for example, if a particular person appears in the CCTV field of view an unusual number of times in a short time frame.
Up Next: Our next article will examine another biometric technology of the future: Earlobe Recognition.
Ravi Das is a Cybersecurity Consultant and Business Development Specialist. He also does Cybersecurity Consulting through his private practice, RaviDas Tech, Inc. He also possesses the Certified in Cybersecurity (CC) cert from the ISC2.
Visit his website at mltechnologies.io
