Formålet med projektet er at udvide det videnskabelige grundlag for maskinlæringsmetoder så de kan bruges i adaptive akustiske systemer. Systemet, der undersøges, er højttalere i et rum og de adaptive faktorer er dels de fysiske forhold i rummet og dels brugerens præferencer for god lydkvalitet som funktion af konteksten. Målet er en markant forøgelse af brugertilfredsheden for et fuldt adaptivt system sammenlignet med et konventionelt system.

Long-Term Electroencephalographic Monitoring (LTVEM) is used for more precise diagnostic information to optimize treatment and care for epileptic patients. LTVEM mainly takes place in hospitals and patients can be referred for LTVEM for 24 hours or up to several days. LTVEM produces a vast amount of EEG data that must be reviewed and validated – largely a manual procedure today that can take 6-8 hours for each patient, using time of technicians and neurologists. We are developing cNode, a device that integrates with existing LTVEM equipment to avoid the need for manual analysis of the EEG data. The firsts step in the cNode process converts the EEG signal into spikes, similar to the way of communication of the brain. The spike data is fed into a Spike Neural Network (SNN) with on-chip learning and a unique pattern recognition ability. The output is the seizure onset and timestamp. The cNode prototype is integrated on an electronic board and its functionality and ability to detect seizure has been confirmed in various in-vitro and human in-silico models