A Wireless Animal Robot Stimulation System Based on Neuronal Electrical Signal Characteristics

As a stimulus signal, coded electrical signals can control the motion behavior of animals, which has been widely used in the field of animal robots. In current research, most of the stimulus signals used by researchers are traditional waveforms, such as square waves. To enrich the stimulus waveform, a wireless animal robot stimulation system based on neuronal electrical signal characteristics is presented in this paper. The stimulator uses the CC1101 wireless module to control animal behavior through brain stimulation. The LabVIEW-based graphical user interface(GUI) can manipulate brain stimulation remotely while the stimulator powered by battery. Additionally, The spikes of animals have been simulated by this system through Direct Digital Synthesizer(DDS) algorithm. The GUI enable users to customize the combination of these analog spike signals. The recombined signals are sent to the stimulator through CC1101 as stimulus signals. In vivo experiments conducted on five pigeons verified the efficacy of the stimulation mechanism. The analog spike signal with an amplitude of 3-5V successfully caused the pigeon’s turning behavior. The feasibility of the analog spike signals as stimulus signals was successfully verifified. Increased the diversity of stimulus waveforms in the field of animal robots.

An FPGA Implementation of Deep Spiking Neural Networks for Low-Power and Fast Classification

A spiking neural network (SNN) is a type of biological plausibility model that performs information processing based on spikes. Training a deep SNN effectively is challenging due to the nondifferention of spike signals. Recent advances have shown that high-performance SNNs can be obtained by converting convolutional neural networks (CNNs). However, the large-scale SNNs are poorly served by conventional architectures due to the dynamic nature of spiking neurons. In this letter, we propose a hardware architecture to enable efficient implementation of SNNs. All layers in the network are mapped on one chip so that the computation of different time steps can be done in parallel to reduce latency. We propose new spiking max-pooling method to reduce computation complexity. In addition, we apply approaches based on shift register and coarsely grained parallels to accelerate convolution operation. We also investigate the effect of different encoding methods on SNN accuracy. Finally, we validate the hardware architecture on the Xilinx Zynq ZCU102. The experimental results on the MNIST data set show that it can achieve an accuracy of 98.94% with eight-bit quantized weights. Furthermore, it achieves 164 frames per second (FPS) under 150 MHz clock frequency and obtains 41[Formula: see text] speed-up compared to CPU implementation and 22 times lower power than GPU implementation.

Complex spikes encode reward expectation signals during visuomotor association learning

Climbing fiber input to Purkinje cells has been thought to instruct learning related changes in simple spikes and cause behavioral changes through an error-based learning mechanism. Although, this framework explains simple motor learning, it cannot be extended to learning higher-order skills. Recently the cerebellum has been implicated in a variety of cognitive tasks and reward-based learning. Here we show that when a monkey learns a new visuomotor association, complex spikes predict the time of the beginning of the trial in a learning independent manner as well as encode a learning contingent reward expectation signal after the stimulus onset and reward delivery. These complex spike signals are unrelated to and were unlikely to instruct the reward based signal found in the simple spikes. Our results provide a more general role of complex spikes in learning and higher-order processing while gathering evidence for their participation in reward based learning.

Ultracompact Multielectrode Array for Neurological Monitoring

Patients with paralysis, spinal cord injury, or amputated limbs could benefit from using brain–machine interface technology for communication and neurorehabilitation. In this study, a 32-channel three-dimensional (3D) multielectrode probe array was developed for the neural interface system of a brain–machine interface to monitor neural activity. A novel microassembly technique involving lead transfer was used to prevent misalignment in the bonding plane during the orthogonal assembly of the 3D multielectrode probe array. Standard microassembly and biopackaging processes were utilized to implement the proposed lead transfer technique. The maximum profile of the integrated 3D neural device was set to 0.50 mm above the pia mater to reduce trauma to brain cells. Benchtop tests characterized the electrical impedance of the neural device. A characterization test revealed that the impedance of the 3D multielectrode probe array was on average approximately 0.55 MΩ at a frequency of 1 KHz. Moreover, in vitro cytotoxicity tests verified the biocompatibility of the device. Subsequently, 3D multielectrode probe arrays were implanted in rats and exhibited the capability to record local field potentials and spike signals.

Hyperacuity Bayesian methods to enhance temporal resolution of two-photon recording of the complex spikes in the cerebellar Purkinje cells

Two-photon imaging is a major recording technique in neuroscience but its low sampling rate imposes a severe limit of elucidating high temporal profiles of neuronal dynamics. Here we developed two hyperacuity Bayesian algorithms to improve spike detection and spike time precision, minimizing the estimation error supervised by the ground-truth given as the electrical spike signals. The benchmark showed that our algorithms outperformed other unsupervised algorithms maximizing the likelihood of the estimates for both experimental and simulation data. We argue that the supervised algorithms are useful tools to improve spike estimation of two-photon recording in case ground truth signals are available.