Vertically integrated spiking cone photoreceptor arrays for color perception

Area of Science:

• Neuromorphic Engineering
• Photoreceptor Technology
• Artificial Vision

Background:

• Biological cone photoreceptors enable energy-efficient color vision through selective light transduction and spiking representations.
• Developing artificial devices with both color selectivity and spike-encoding capabilities remains a significant challenge in visual sensing technology.

Purpose of the Study:

• To propose and demonstrate a metal oxide-based vertically integrated spiking cone photoreceptor array.
• To achieve direct transduction of light wavelengths into spike trains with high energy efficiency.
• To enable color image recognition and discrimination using artificial cone photoreceptors.

Main Methods:

• Fabrication of a metal oxide-based vertically integrated spiking cone photoreceptor array.
• Utilizing lights of three specific wavelengths as pseudo-primary colors for image input.
• Evaluating the device's performance in color image recognition and mixed color discrimination tasks.

Main Results:

• The developed spiking cone photoreceptors directly transduce light wavelengths into spike trains.
• Achieved ultralow power consumption of less than 400 picowatts per spike in visible light.
• Demonstrated improved accuracy in recognizing mixed colors and 'colorful' images.

Conclusions:

• The proposed device mimics biological cone photoreceptors' function with high energy efficiency.
• This technology enables hardware spiking neural networks with biologically plausible visual perception.
• Offers significant potential for developing advanced dynamic vision sensors and neuromorphic systems.