With increased age, challenges like cognitive and physical decline are more likely to occur and involve the age-related frailty syndrome, which is characterized by lower resistance against stress, diseases, and other exogenous influences. Frail older adults have an enhanced need of health care utilization, are more likely to fall and be hospitalized due to reduced cognitive and physical capabilities. Due to correlation between frailty and physical fitness, there are physical performance tests, to assess a person's mobility, and the associated risk to become frail. In this thesis, the fitness level is automatically assessed based on measurement of gait parameters and sit-to-stand performance. The first measurement method analyses human gait for walking velocity, distance, and duration. Further, a literature-based scale-space filtering approach is compared to a machine learning approach to determine gait events, which are further used to estimate gait cycle components. The comparison shows better results for the machine-learning approach, and gait cycle component estimation comparable to state-of-the-art measurements using wearable accelerometers. The second measurement method is split into two problems: the detection of sit-to-stand movements and the measurement of sit-to-stand transition durations using curve-sketching. Stand-up detection achieves best results using a random-forest algorithm, the duration measurement achieves results comparable to state-of-the-art methods. For both measurements, person tracking data from depth sensors are utilized, which allows non-intrusive, privacy protecting capturing of motion at habitual speed in the homes of older persons. For evaluation of the proposed measurement approaches, two datasets have been recorded and manually annotated. A holistic system combining the measurement algorithms was tested in an 8-week field trial with 4 older adults.