Navigation auf uzh.ch
Progress in bio- and neural- recording technologies has made it possible recently to record, amplify, and process hundreds of bio-signals directly at the front-end, for example next to large micro-electrode arrays implanted in different brain areas. However, most of the bio-signal recording systems developed to date are mainly limited to signal conditioning aspects (e.g., amplifying and filtering) and analog-to-digital conversion aspects, with wireless transmission of the raw data to external data analysis, storage, and computing systems.
We are developing low-power circuits for converting these signals into asynchronous "Address-Events" (like spikes), and low-power spiking neural networks that can learn to optimally decode and interpret them. The possibility of integrating these circuits on the same chip that integrates the bio-signal recording circuits will allow us to develop intelligent bio-electronics interfaces (e.g., intelligent Brain-Machine Interfaces) that can both record and interpret the data locally, without needing to spend energy for transmitting large amounts of information to off-chip processing modules.