Light source color coordinate estimation system and deep learning method thereof

A deep learning method of a light source color coordinate estimation system, comprising:

• providing a plurality of light sources that are different from each other;
• and sequentially executing a deep learning process for the light sources;
• wherein the deep learning process executing for each of the light sources includes:
• receiving, by a plurality of photo detectors having different detection wavebands respectively, a beam from the light source;
• wherein a number of the plurality of photo detectors is larger than three;
• generating, by the plurality of photo detectors, spectral responses respectively;
• obtaining energy integral values corresponding to the detection wavebands for each of the spectral responses through a normalization calculation circuit;
• dividing the energy integral values by a largest of the energy integral values respectively through the normalization calculation circuit so as to generate a plurality of normalized energy integral values;
• receiving the plurality of normalized energy integral values through a neural network, wherein the neural network includes activity functions and weights;
• converting the plurality of normalized energy integral values into an estimated color coordinate of the light source by calculations of the activity functions and the weights;
• calculating a color coordinate deviation between the estimated color coordinate and a preset color coordinate;
• and
• adjusting at least one of the weights according to a back propagation algorithm and the color coordinate deviation.