High data-rate ultra low-power backscatter wireless communication systems for brain-computer interfaces

High data-rate ultra low-power backscatter wireless communication systems for brain-computer interfaces Eleftherios Kampianakis Chair of the Supervisory Committee: Professor Matthew S. Reynolds Department of Electrical & Computer Engineering Neural interfacing is a promising technology for effectively treating a multitude of challenging clinical conditions. Recent research has demonstrated that some tetraplegic patients can control robotic limbs using a brain-computer interface (BCI), signifying the beginning of an era wherein many forms of paralysis may be treatable with a neuro-prosthesis. However, the current state of the art is bulky, tethered, and impractical for applications outside a clinical lab setting. Moreover, current wireless communication approaches for brain computer interfaces (BCI) do not meet the necessary specifications for power, size, and bandwidth. In contrast, we developed fully integrated BCIs equipped with high data-rate and low power miniaturized wireless backscatter communication systems to enable the development of autonomous brain-controlled prosthetics. First, we proposed a wireless μ-Power, low-noise frequency mixing approach for extending the passband frequency response of existing neural interfaces. We demonstrated the translation of a pre-recorded mouse electrocorticogram from a frequency range of 0.5 Hz to 100 Hz up to an intermediate frequency (IF) of 407 Hz, thus enabling the use of an existing integrated circuit (IC) for electrocorticography (ECoG), despite its low-frequency cutoff of 12 Hz. Subsequently, we presented a dual-band implantable BCI that integrates 47%-efficient high frequency (HF) wireless power delivery into a 5 Mb/s ultra-high frequency (UHF) backscatter communication. The implant system supports ten neural channels sampled at 26.10 kHz and four electromyography (EMG) channels sampled at 1.628 kHz and can communicate with a custom software-defined-radio-based external system with a packet error ratio (PER) that is better than 0.19 % at an implant depth of up to 3 cm. Finally, in order to enable neural plasticity experiments inside the home cages of freely behaving animals, we developed a 25 Mbps backscatter-based data uplink for Neurochip 3 using a differential quadrature phase shift keying (DQPSK) constellation. We statically collected 104 packets from 126 locations, and the system exhibited effectively 0 % (PER) for all but two of the surveyed sites despite the reverberant cavity effects of the animal cage that critically impair the communication channel.