2021 BRN Discussion, page-17395

A search for google spiking neural network patents/applications got 6 hits, but they were mostly for cloud-based systems except for WO2020167314A1 which uses IR LiDAR in a mobile phone.

https://worldwide.espacenet.com/patent/search/family/065685974/publication/WO2020167314A1?q=WO2020167314A1

1. A method comprising: obtaining, by a computing system, an infrared image of an object;determining, by the computing system, a difference of Gaussian image that represents features of the infrared image that have spatial frequencies within a spatial frequency range defined by a first Gaussian operator and a second Gaussian operator;identifying, by the computing system, one or more blob regions within the difference of Gaussian image, wherein each blob region of the one or more blob regions comprises a region of connected pixels in die difference of Gaussian image; andbased on identify ing the one or more blob regions within the difference of Gaussian image, determining, by the computing system, that the infrared image represents the object illuminated by a pattern projected onto the object by an infrared projector.








The specification does not describe the neuron circuit layout for a SNN, but only refers to SNNs as a "me too". :


[119] In some embodiments, ANN 600 may be or may include therein aspects of a convolutional artificial neural network (CANN). For example, ANN 600 may include pooling layers (i.e., downsampling layers) between layers 610, 620, 630, and 640. Further, ANN 600 may additionally include aspects of probabilistic neural networks, time-delay neural networks, regulatory feedback neural networks, and spiking neural networks, among other types of neural networks not herein discussed.

So that leaves open the possibility that Google are licensing in their SNN IP.

Of course we need to keep in mind the 18 month lacuna before a patent application is published, but could they really design something to get around our patents in that time?