.png){kind=icon}
.png){kind=icon}
Return
Compressing Ultra-Dense Neural Recordings in Space and Time: A 3-D Modeling Approach
.png){kind=icon}
OA
AI
Y
G
S
Z
X
M
DOI：10.1109/tnsre.2026.3716012.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Increasing the number of channels is essential for improving neural signal acquisition in wireless implantable brain-computer interfaces (iBCIs). However, more channels raise power consumption and data rates, placing additional demands on the limited energy supply and communication bandwidth of implanted devices. A central challenge is how to fully leverage the global spatiotemporal correlations within neural signals to further enhance compression efficiency while preserving the integrity of critical information. This paper proposes a three-dimensional (3D) spatiotemporal neural signal compression method based on compressed sensing for ultra-high-channel wireless neural recording systems. This approach comprises two components: a 3D spatiotemporal matrix representation of neural signals and three-dimensional compressed sensing (3D-CS). Unlike traditional methods that focus solely on spatial correlations among multi-channel neural signals, this approach exploits the spatiotemporal correlations inherent in ultra-high-channel neural signals. It represents neural signals as 3D spatiotemporal matrices while synchronizing action potentials (APs) and local field potentials (LFPs), enabling holistic compression. The method was validated using real-world data from a 1024-channel system. Results showed that after compression with this method, all APs remained intact, with all LFPs achieving a structural similarity index measure (SSIM) greater than 0.95. The average signal-to-noise and distortion ratio (SDNR) reached approximately 24.16 dB, achieving a total compression ratio (CR) of 156. The findings demonstrate that this approach provides an efficient, low-power data compression solution for ultra-high-channel wireless implantable brain-computer interfaces.
Keywords:
Implantable brain-computer interface
data compression
3D spatiotemporal matrix
3D-CS
Journal
IF:
5.2
Papers:
448
Citations:
1.6W
Organization
Cited Papers