Detecting human presence in an outdoor monitored site
AssigneeProject Canary, Pbc
InventorRay K. Mistry
Life of this patent
Estimating the term…
Abstract
Methods and systems for detecting human presence in an outdoor monitored site are disclosed. In an configuration, the method may include providing at least three transmitters for transmitting radio signals and at least three receivers for receiving radio signals. Channel state information (CSI) data is extracted from the radio signals. The CSI data include phase values and amplitude values. The CSI data is converted into an image representation map and then a surface map is generated using Machine Learning (ML) networks. The surface map is a representation of correspondence between the image representation map and a 3-dimensional (3D) surface associated with a human body.
Claims22
A method of detecting human presence in an outdoor monitored site, the method comprising:
- providing a first air quality monitor comprising:
- a first transmitter for transmitting radio signals;
- and
- a first receiver for receiving radio signals;
- providing a second air quality monitor comprising:
- a second transmitter for transmitting radio signals;
- and
- a second receiver for receiving radio signals;
- providing a third air quality monitor comprising:
- a third transmitter for transmitting radio signals;
- and
- a third receiver for receiving radio signals;
- wherein each of the first receiver, the second receiver, and the third receiver is configured to receive radio signals transmitted by the first transmitter, the second transmitter, and the third transmitter;
- receiving, from the first receiver, the second receiver, and the third receiver, radio signals as received by each of the first receiver, the second receiver, and the third receiver, corresponding to a space through which the radio signals propagate;
- extracting, by a first machine learning network, channel state information data from the radio signals;
- wherein the channel state information data comprises:
- phase values;
- and
- amplitude values;
- converting, by the first machine learning network, the channel state information data into an image representation map of a predefined configuration, by performing domain translation of the channel state information data using a two-branch encoder-decoder network;
- and
- generating, by a second machine learning network, a surface map based on the image representation map,
- wherein the surface map is a representation of correspondence between the image representation map and a 3-dimensional (3D) surface associated with a human body.
A system for detecting human presence in an outdoor monitored site, the system comprising:
- at least three transmitters for transmitting radio signals;
- at least three receivers for receiving radio signals;
- wherein each of the at least three transmitters is configured to receive radio signals transmitted by each of the at least three receivers;
- and
- a human presence detecting device comprising:
- a processor;
- and
- a memory storing a plurality of processor-executable instructions which, upon execution by the processor, cause the processor to:
- extract, by a first machine learning network, channel state information data from the radio signals, wherein the channel state information data comprises:
- phase values;
- and
- amplitude values;
- convert, by the first machine learning network, the channel state information data into an image representation map, by performing domain translation of the channel state information data using a two-branch encoder-decoder network;
- and
- generate, by a second machine learning network, a surface map based on the image representation map, wherein the surface map is a representation of correspondence between the image representation map and a 3-dimensional (3D) surface associated with a human body.
A method of detecting human presence in an outdoor monitored site, the method comprising:
- receiving from a horizontal antenna array, a plurality of two-dimensional (2D) horizontal heatmaps, wherein each of the plurality of two-dimensional (2D) horizontal heatmaps is a projection of RF signal reflections on a plane parallel to a ground associated with the outdoor monitored site;
- receiving from a vertical antenna array, a plurality of two-dimensional (2D) vertical heatmaps, wherein each of the plurality of two-dimensional (2D) vertical heatmaps is a projection of RF signal reflections on a plane perpendicular to the ground associated with the outdoor monitored site;
- inputting the plurality of two-dimensional (2D) horizontal heatmaps and the plurality of two-dimensional (2D) vertical heatmaps to a student network of a machine learning network;
- inputting image data synchronized with the plurality of two-dimensional (2D) horizontal heatmaps and the plurality of two-dimensional (2D) horizontal heatmaps to a teacher network of a machine learning network, wherein the teacher network is to provide cross-modal supervision for the student network;
- and
- receiving, from the student network, predicted key-point confidence maps, corresponding to one or more anatomical parts of a human body.
Dependencies are inferred from the claim text.