Artificial retinal prosthesis

An artificial retinal prosthesis, comprising:

• a pixel array comprising a plurality of pixels, each of the pixels comprising a photosensor to receive image, an electrode-driving circuitry coupled to the photosensor, and an electrode driven by the electrode-driving circuitry configured to stimulate target neuron cells;
• and
• a processing unit, coupled to the pixel array and stored a plurality of convolutional matrices, each of the convolutional matrices being corresponding to spatial size of each of the pixels and spanning close to a receptive field of target neuron cells such that the target neuron cells are electrically stimulated, which are configured to determine electrical stimulation output by each electrode to induce perception of vision, the processing unit comprising an analog-to-digital converter (ADC), a digital core coupled with the ADC and a digital-to-analog converter (DAC) coupled with the digital core;
• wherein the ADC converts an image received by the photosensors in the pixel array to a binary form;
• wherein the digital core calculates and obtains an electrical stimulation waveform for each of the pixels based on the following equation:
• Stim(i,j,t)=HANN(Img(i,j,t))
• wherein Stim (i, j, t) represents the electrical stimulation waveform for the pixel at location (i, j) at time (t), Img (i, j, t) represents the image received by the photosensor of the pixel at location (i, j) at time (t), HANN is a processing function of an artificial neural net (ANN) involving a calculation of a weight parameter based on at least one convolutional layer with a convolution matrix of corresponding to a spatial size covering the receptive field for normal retinal ganglion cell and electrical stimulation waveforms calculated at previous times and the present time (t);
• and
• wherein the DAC converts the electrical stimulation waveform from a digital format to an analog format, and outputs the electrical stimulation waveform in the analog format to the electrode-driving circuitry to drive electrode that stimulates the target neuron cells.

An artificial retinal prosthesis, comprising:

• a pixel array comprising a plurality of pixels, each of the pixels comprising a photosensor to receive image, an electrode-driving circuitry coupled to the photosensor, and an electrode driven by the electrode-driving circuitry configured to stimulate target neuron cells;
• and
• a processing unit, coupled to the pixel array and stored a plurality of convolutional matrices, each of the convolutional matrices being corresponding to spatial size of each of the pixels and spanning close to a receptive field of target neuron cells such that the target neuron cells are electrically stimulated, which are configured to determine electrical stimulation output by each electrode to induce perception of vision, the processing unit comprising an analog-to-digital converter (ADC), a digital core coupled with the ADC and a digital-to-analog converter (DAC) coupled with the digital core;
• wherein the ADC converts an image received by the photosensors in the pixel array to a binary form;
• wherein the digital core calculates and obtains an electrical stimulation waveform for each of the pixels based on the following equation:
• Stim(i,j,t)=HANN(Img(i,j,t))
• wherein Stim (i, j, t) represents the electrical stimulation waveform for the pixel at location (i, j) at time (t), Img (i, j, t) represents the image received by the photosensor of the pixel at location (i, j) at time (t), HANN is a processing function of an artificial neural net (ANN) involving a calculation of a weight parameter based on at least one convolutional layer with a convolution matrix of corresponding to a spatial size covering the receptive field for normal retinal ganglion cell and at least one recurrent neural network from previous time step X(t−1) combined with current inputs Input (t) (a tensor) of the recurrent neural network to generate the state tensor X(t) at time (t);
• and
• wherein the DAC converts the electrical stimulation waveform from a digital format to an analog format, and outputs the electrical stimulation waveform in the analog format to the electrode-driving circuitry to drive the electrode that stimulates the target neuron cells.