A warning sign of frailty is imbalance. Psycho-motor therapists run tests to evaluate the balance deterioration but not often enough to track the rapidly changing condition of the elderly. The proposed system collects fine-grained data from a smart cane and processes them with Machine Learning (ML) techniques. The originality of our proposition lies in its personalization by the elderly biomarkers in ML algorithms. Our experiments indicate that we can observe the orientation of locomotion through the cane as well as recognize characteristics of specific participants ambulation in the uncontrolled scenario.