FPGA implementation of deep-learning recurrent neural networks with sub-millisecond real-time latency for BCI-decoding of large-scale neural sensors (104 nodes).

Advances in neurotechnology are expected to provide access to thousands of neural channel recordings including neuronal spiking, multiunit activity and local field potentials. In addition, recent studies have shown that deep learning, in particular recurrent neural networks (RNNs), provide promising approaches for decoding of large-scale neural data. These approaches involve computationally intensive algorithms with millions of parameters. In this context, an important challenge in the application of neural decoding to next generation brain-computer interfaces for complex human tasks is the development of low-latency real-time implementations. We demonstrate a Field-Programmable Gate Array (FPGA) implementation of Long Short-Term Memory (LSTM) RNNs for decoding 10,000 channels of neural data on a mobile lowpower embedded system platform called "NeuroCoder". We provide a proof of concept in the context of decoding 20dimensional spectrotemporal representation of spoken words from simulated 10,000 neural channels. In this particular case, the LSTM model included 4,042,420 parameters. In addition to providing multiple communication interfaces for the BCI system, the NeuroCoder platform can achieve sub-millisecond real-time latencies.

Picosecond ultrasonic measurements using an optical mask.

In this paper we describe results obtained using a variation of the picosecond ultrasonics technique. We place a transparent optical mask very close to the surface of the sample. The lower surface of the mask has a series of grooves to produce a variation of the intensity of the pump and probe light pulses across the surface of the sample. Because the light intensity varies with position, the application of the pump light pulse can generate surface acoustic waves with a wavelength equal to the period of the mask. We report results obtained in this way and discuss the possible practical applications of this new approach.

A scanning acoustic microscope based on picosecond ultrasonics.

We report on the development of a new type of scanning acoustic microscope. We use a femtosecond light pulse to generate a short sound pulse, and then focus this sound onto the sample by means of a specially designed and microfabricated acoustic lens of radius a few microns. The sound travels to the sample through a thin layer of water. The sound reflected from the sample is collected by the lens and then passes through a monolithically integrated optical resonant cavity. The induced change in the properties of this cavity are measured using a time-delayed probe light pulse. We describe some of the challenges involved in the construction and operation of this high-precision metrology apparatus and present some preliminary results.

Active microelectronic neurosensor arrays for implantable brain communication interfaces.

We have built a wireless implantable microelectronic device for transmitting cortical signals transcutaneously. The device is aimed at interfacing a cortical microelectrode array to an external computer for neural control applications. Our implantable microsystem enables 16-channel broadband neural recording in a nonhuman primate brain by converting these signals to a digital stream of infrared light pulses for transmission through the skin. The implantable unit employs a flexible polymer substrate onto which we have integrated ultra-low power amplification with analog multiplexing, an analog-to-digital converter, a low power digital controller chip, and infrared telemetry. The scalable 16-channel microsystem can employ any of several modalities of power supply, including radio frequency by induction, or infrared light via photovoltaic conversion. As of the time of this report, the implant has been tested as a subchronic unit in nonhuman primates ( approximately 1 month), yielding robust spike and broadband neural data on all available channels.

Highly efficient resonant coupling of optical excitations in hybrid organic/inorganic semiconductor nanostructures.

The integration of organic and inorganic semiconductors on the nanoscale offers the possibility of developing new photonic devices that combine the best features of these two distinct classes of material. Such devices could, for example, benefit from the large oscillator strengths found in organic materials and the nonlinear optical properties of inorganic species. Here we describe a novel hybrid organic/inorganic nanocomposite in which alternating monolayers of J-aggregates of cyanine dye and crystalline semiconductor quantum dots are grown by a layer-by-layer self-assembly technique. We demonstrate near-field photon-mediated coupling of vastly dissimilar optical excitations in the two materials that can reach efficiencies of up to 98% at room temperature. By varying the size of the quantum dots and thus tuning their optical resonance for absorption and emission, we also show how the ability of J-aggregates to harvest light can be harnessed to increase the effective absorption cross section of the quantum dots by up to a factor of ten. Combining organic and inorganic semiconductors in this way could lead to novel nanoscale designs for light-emitting, photovoltaic and sensor applications.

Strong coupling of nonlinear electronic and biological oscillators: reaching the "amplitude death" regime.

Interaction between an electronic and a biological circuit has been investigated for a pair of electrically connected nonlinear oscillators, with a spontaneously oscillating olivary neuron as the single-cell biological element. By varying the coupling strength between the oscillators, we observe a range of behaviors predicted by model calculations, including a reversible low-energy dissipation "amplitude death" where the oscillations in the coupled system cease entirely.

Development of an integrated microelectrode/microelectronic device for brain implantable neuroengineering applications.

An ultra-low power analog CMOS chip and a silicon based microelectrode array have been fully integrated to a microminiaturized "neuroport" for brain implantable neuroengineering applications. The CMOS IC included preamplifier and multiplexing circuitry, and a hybrid flip-chip bonding technique was developed to fabricate a functional , encapsulated microminiaturized neuroprobe device. As a proof-of-concept demonstration, we have measured local field potentials from thalamocortical brain slices of rats, suggesting that the new neuroport can form a prime platform for the development of a microminiaturized neural interface to the brain in a single implantable unit.

Photovoltaic energy converter as a chipscale high efficiency power source for implanted active microelectronic devices.

We report the development of a microscale photovoltaic energy converter which has been designed and implemented to deliver power to CMOS-based microelectronic chips. The design targets the delivery of voltages on the order of 3V with power levels in excess of 10 mW. The geometry of the prototype device, which has been fabricated and tested, is specifically designed for coupling to an optical fiber, to facilitate remote power delivery in implantable component environment.

Coherent magnetization rotation and phase control by ultrashort optical pulses in CrO(2) thin films.

We have applied photoexcitation by ultrashort laser pulses to single crystal thin CrO(2) films to trigger coherent transient magnetization rotation on a subnanosecond time scale, in macroscale single domains. Moreover, by applying the photoexcitation by pairs of temporally separated pump pulses, the transient precession of the magnetization can be phase controlled, depending on the time separation between the pulses. The mechanism behind the photoexcitation originates from the modulation of the magnetocrystalline anisotropy by nonthermal hot electron spins.

Pulsed room-temperature operation of a blue-green ZnSe quantum-well diode laser.

Low-threshold current density operation has been obtained from (Zn, Cd)Se/Zn(S, Se) quantum-well diode lasers in the blue-green spectrum. The devices, with reflective facet coatings, have been operated at room temperature under pulsed conditions with threshold current densities of I(th) approximately 1 kA/cm(2).

Fundamental absorption edge in laser window CdTe.

Fundamental absorption edge and the ir absorption at 10.6-microm wavelengths have been studied in CdTe laser window samples of varying optical quality. The results show that the edge spectroscopy can be used as a sensitive diagnostic technique for evaluating the over-all optical quality of the material.