Currently a questionnaire-based method known as ECOG, assigning a score according to the patient's response is employed for tracking cancer patient's physical status during chemotherapy. This approach is quite subjective, and prone to error and inaccuracy. Walking test (WT) has been used in the past as an objective physical assessment of patients with cardiopulmonary diseases and to very limited extent has been experimented on oncology patients. In the conventional WT, only the distance travelled by the patient is measured without taking into account the characteristics of the gait and posture that reflects a patient's fitness. Towards developing an objective method to track cancer patient's physical status, inertial sensors are used to measure walk and posture characteristics during a WT. Using a set of 17 sensors, the inertial signals corresponding to position, velocity, acceleration, orientation, angular velocity and angular acceleration are recorded based on a 23 degree of freedom humanoid model. The data streams obtained are subsequently segmented by an intrinsic clustering algorithm known as Minimum Message Length encoding (MML) forming a Gaussian Mixture Model (GMM). Several postural states (exemplar motion primitives) are captured in the resultant model, which is subsequently utilized to derive a holistic index corresponding to the physical ambulatory status of patient. The proposed method is applied to two sets of experimental data both pre- and post-chemotherapy conditions. The simulation scenario for chemotherapy patients and typical gait behavior has been devised in consultation with the collaborating Oncologist. The results are encouraging as they clearly distinguish between different simulated fitness conditions and can potentially assist oncologist to make a decision about continuation or termination of treatment.