Patients with Alzheimers disease show cognitive decline commonly associated with psycho-behavioural disorders like depression, apathy and motor behaviour disturbances. However current evaluations of psycho-behavioural disorders are based on interviews and battery of neuropsychological tests with the presence of a clinician. So these evaluations show limits of subjectivity (e.g., subjective interpretation of clinician at a date t). In this work, we study the ability of a proposed automatic video activity recognition system to detect activity changes between elderly subjects with and without dementia during a clinical experimentation. A total of 28 volunteers (11 healthy elderly subjects, 17 Alzheimer's disease patients (AD)) participate to the experimentation. The proposed study shows that we could differentiate the two profiles of participants based on motor activity parameters, such as the walking speed, computed from the proposed automatic video activity recognition system. These primary results are promising and validating the interest of automatic analysis of video as an objective evaluation tool providing comparative results between participants and over the time.