A particular way or manner of moving on foot is a definition for gait. Every person has his or her own way of walking. From early medical studies it appears that there are twenty-four different components to human gait, and that if all the measurements are considered, gait is unique. This has made gait recognition an interesting topic to be used for identifying individuals by the manner in which they walk. Figure 1 illustrates the complex biological process of the musculo-skeletal system, which can be divided into several types of sub events of human-gait. The instances that are shown in this figure are used to extract parameters for being used as an identification system of each individual.
The analysis of biometric gait recognition has been studied for a longer period of time for the use in identification, surveillance and forensic systems and is becoming important, since it can provide more reliable and efficient means of identity verification.
Figure 1: Division of the gait cycle into five stance phase periods and two swing phase periods
Today, computer systems demand authentication in case of using the system. Typically, the authentication is performed at login time either with a password, token, biometric characteristic and/or a combination of these. Performing the last mentioned might give further guarantee that the claimed user logging in is the authorized user instead of a burglar. However, once the user has been granted access; most systems assume that the user is continuously legitimated into the system.
In critical or high security environments, it should be ensured that the user must be the legitimated throughout usage. Therefore, user authentication needs to be performed in a continuous way within the time the system is actively being used. Furthermore, authentication needs to be “attractive” for the user. This means that in the authentication process the users do not need to do anything special, like for example periodically entering a password. Continuous authentication using biometrics can fit these needs. Thus, one of the important requirements in continuous authentication is unobtrusiveness, since this can be monitored in a non-intrusive way. The Wearable Sensor (WS) based method can be a very good candidate to fulfill this requirement, compared to current knowledge-based mechanisms.
Gait-Recognition Tecniques
There are three different approaches in gait recognition: Machine Vision Based, Floor Sensor Based and Wearable Sensor Based Gait Recognition. These will be explained in the next paragraphs.
Machine Vision (MV) Based
In the machine vision approaches, the system will typically consist of several digital or analog cameras with suitable optics for acquiring the gait data. Techniques such as thresholding to convert images into black and white; pixel counting to count the number of light or dark pixels; or background segmentation are used to extract features to identify a person. Figure 2 shows an example of the MV-based approach with processed background segmentation.
Figure 2: Background segmentation for extracting the silhouette picture (subtraction).
Earlier gait-recognition studies have shown promising results. An experiment with 1870 gait datasets from 122 subjects reported an recognition rate of 78% in an identification scenario. This was further improved to a rate of 90% by other research. Most of the current gait recognition approaches are MV-based. The main advantage for this type of recognition compared to other biometric systems is that persons are captured unobtrusively from a distance. Even though MV-based gait analysis is not that precise as other biometrics, e.g. fingerprints, it is still useful for surveillance scenarios.
Floor Sensor (FS) Based
In the floor sensor approach the sensors are placed on a mat along the floor which makes these methods suitable for controlling access to buildings. When people walk across the mat, the force to the ground is measured, this is also known as the GRF (Ground Reaction Force). In a research from the University of Southampton, such a floor sensor for gait recognition was prototyped and is illustrated in Figure 3.
Figure 3: Gait collection by floor sensors. a) shows foot steps recognized, b) shows the time spent at each location in a), c) shows footstep profiles for heel and toe strikes, and finally d) is a picture of a prototype floor sensor carpet
Their experiment had 15 subjects and three different features were extracted, namely the stride length (the distance traveled by the heel of one foot to the next time the same foot strikes down), stride cadence (the rhythm of a person’s walk) and TOH ratio (the time on toe to the time on heel ratio). Using the TOH ratio an recognition rate of 80% could be achieved. Different studies with small number of test persons (10 – 15) exist which report recognition rates up to 98.2. Jenkins and Ellis had 62 test persons and only reported a recognition rate of 39%.
Wearable Sensor (WS) Based
The wearable sensor recognition approach is the newest gait-recognition among the other mentioned earlier. This is based on wearing motion recording sensors on the body (see Figure 4) of the person in different places; on the waist, pockets, shoes and so forth.
The wearable sensors can be accelerometers (measure acceleration), gyro sensors (measure rotation and number of degrees per second of rotation), force sensors (measure the force when walking) etc. The main advantage of WS-based gait recognition is that it provides an unobtrusive authentication method for mobile devices containing accelerometers (like mobile phones, PDAs etc.). Therefore, it can be applied for continuous verification of the identity of the user without his/her intervention. This is a great advantage to other biometric systems like fingerprint or face recognition which are also suitable for implementation on mobile phones but require active user intervention. This advantage of accelerometer based gait recognition compensates the so far worse recognition rates.
[Site is under construction - 21/02-11]