Robust Volume Control Personalisation from on-Line Preference Feedback

We describe a learning volume control (LVC) algorithm that learns the volume control operations of a hearing aid user in such a way that the internal settings of the volume control will over time absorb the user´s preferences. For practical use, the algorithm should be robust against inconsistent behaviour of the user. We propose two algorithms for LVC (based on LMS learning and Kalman filtering) and demonstrate desirable properties of our LVC in simulations. Further, we provide evidence for the practical relevance of our algorithms in a listening test.