Balance and gait rehabilitation are critical to the recovery of motor function in hemiparetic stroke patients. Musical biofeedback has been shown to hold great rehabilitative potential due to its emotional appeal, capacity to induce and facilitate bodily movement and documented therapeutic benefits. This thesis aimed to investigate the types of user-tailored musical biofeedback interactions and strategies most applicable to common rehabilitation protocols. A prototype application based on wireless inertial sensors was built and iteratively evaluated over three development cycles in collaboration with patients and clinicians. Results showed that the developed interactions tailored to static balance, dynamic balance, sit-to-stand and gait may be clinically useful and usable with a number of stroke patient subgroups, promoting autonomy and augmenting conventional training. Future studies must systematically investigate short/long term physical and psychological effects.