A method for human identification using static, activity-specific parameters is presented. This method recovers static body and stride parameters during the gait cycles of humans. Our technique is classified as a gait biometric; however, it does not directly analyze dynamic gait patterns, but uses the action of walking to extract relative body and stride parameters. This approach is an example of an activity-specific biometric: a method of extracting identifying properties of an individual or of an individual's behavior that is applicable only when a person is performing that specific action. To evaluate our parameters, we derive an expected confusion metric in lieu of reporting recognition rates, which are misleading in limited databases. Given a small, yet representative, set of subjects, the expected-confusion metric allows us to predict the identification uncertainty of a feature vector for a larger population of subject. In addition, after multiplying by a dimensional varying scale factor, this transformed expected-confusion gives us the probability of incorrect identification for the feature vector. Last, we test the utility of a variety of body and stride parameters recovered from different viewing conditions and walking speeds, and we use motion-capture data of subjects to discover whether confusion in the parameters is inherently a physical or a visual measurement error property.
展开▼