Abstract: We present a patient-independent, prototype-based spatio-temporal CNN processor for seizure detection, achieving high accuracy without patient-specific data, at an energy of 16.4nJ/class and ...