Contents I. Introduction
The growing recording density of neural interface technologies pave the way towards investigating brain-wide neural dynamics from distributed, layered brain structures [1], [2]. Neuronal population activities can therefore be recorded from thousands of neurons at finer spatial and temporal resolution. In an alternative paradigm, it is also desired to conduct medium density, free-behaving wireless recording experiments [3] in small (e.g., rodents), large (e.g., non-human primates) or flying animal models (e.g., birds, bats, etc.,). With tens of kHz sampling frequency and ≥ 10-bit analog-to-digital conversion resolution, real-time data rate of above-mentioned neural interface are on the order of Gbps and hundreds of Mbps for wired and wireless settings, respectively. In a foreseeable future, the power consumption of the data link technologies for transmitting the neural data will be non-trivial, while telemetry bandwidth of the wireless neural interface is still orders of magnitude lower compared to that of the wireline transceivers. As a result, high fidelity neural signal compression scheme [4]–[6] becomes essential at system-level for saving bandwidth/energy consumption in both wired and wireless neural recording scenarios.
