Gait analysis: Is it easy to learn to walk like someone else?

In this master’s thesis, we will look at whether it is easy or difficult to learn to walk like someone else in such a way that one will be accepted by an authentication system, based on gait. In the last couple of years, there have been some studies about whether gait can be used in order to authenticate a person. If it turns out to be very easy to learn to walk like another person, then gait authentication should probably not be used as the only authentication technique. We investigate the ease of gait mimicking by means of a 3-axis sensor worn by the user. A prototype is created, which reads this acceleration data, and plots it as 3 graphs in a coordinate system, shown on a big screen. The aim is to see whether the user manages to learn to walk in such a way that his graphs match 3 template graphs, which are also plotted in the same coordinate system. Every attempt lasts 5 seconds, and a score between 0 and 100 will be given in the end of each, based on how similar the impersonator’s graphs are to the original graphs. We use Pearson’s correlation to calculate this. The experiment has 13 participants, and we have created 5 different templates which each participant will attempt to imitate 15 times. The results from our data analysis are showing that it actually seems rather easy to learn to walk like another person, and hence to be accepted by a gait authentication system.