
Manufacturer of the Utah Array, the gold-standard electrode system used in the majority of human BCI research.
Open Source
A unified Python framework for spike sorting that wraps multiple sorting algorithms.
Creating the Connexus Direct Data Interface, a high-data-rate BCI for severe motor impairment.
Specializes in CMOS-based high-resolution MEAs with embedded signal processing.
Develops high-density microelectrode arrays (HD-MEA) with FPGA-based real-time spike sorting capabilities.
Develops ultra-low-power mixed-signal neuromorphic processors and sensors for edge AI applications.
Developer of the Akida neuromorphic processor IP and chips.
United States · Company
Produces the industry-standard amplifier chips (RHD/RHS series) used in most neural recording devices.
Real-time spike sorting algorithms are embedded machine learning pipelines that run directly on implantable neural interface hardware, combining adaptive filtering and unsupervised clustering techniques to automatically identify and classify individual neurons' action potentials (spikes) from raw neural recordings in real-time, labeling single-unit activity as it occurs. By processing data on-device rather than transmitting raw signals, these systems dramatically reduce bandwidth needs (since only sorted spike times need to be transmitted, not continuous raw data) and enable closed-loop applications with millisecond latency, where the device can respond to specific neurons' activity almost immediately without the delay of transmitting data to external computers for processing.
This innovation addresses the bandwidth and latency limitations of neural interfaces, where transmitting raw neural data requires enormous bandwidth and processing delays limit responsiveness. By processing on-device, these systems enable more practical and responsive BCIs. Research institutions and companies are developing these technologies.
The technology is essential for high-channel-count neural interfaces and closed-loop applications, where on-device processing becomes necessary. As neural interfaces scale to more channels, on-device processing becomes increasingly important. However, ensuring accuracy, managing computational constraints, and achieving reliable real-time performance remain challenges. The technology represents an important evolution in neural interface capabilities, but requires continued development to achieve the performance and efficiency needed for practical use. Success could enable more practical high-performance BCIs, but the technology must balance processing capabilities with power consumption and device size constraints.