Back to results
US 12,644,956 B2Granted
On-body sensor system and method for automatic interpretation of visual body signals
AssigneeGeorgia Tech Research Corporation
InventorsAshutosh Dhekne, Mostafa H. Ammar, Yifeng Cao
Life of this patent
Estimating the term…
Abstract
An exemplary system and method are disclosed for capturing visual body signals using distance measurements among different parts of the body and for providing classification for them. The exemplary system and method can be employed to generate the classification and provide a second source of communication of the classification to supplement the visual cues provided by such body motion or positioning.
Claims20
A system comprising:
- a plurality of ultrawide-band (UWB) communicating devices configured to be communicatively arranged as a plurality of nodes in a mesh configuration, wherein the plurality of UWB communicating devices includes a first UWB communicating device configured to sequentially transmit in each of a plurality of messaging cycles, a first localization packet of data to each of the other plurality of UWB communicating devices, including a second UWB communicating device, wherein the first UWB communicating device and the second UWB communicating device are configured to determine time associated values corresponding to receipts of the transmitted localization packets sent from the other UWB communicating devices, wherein the determined time associated value is included in the first localization packet, wherein the first UWB communicating device and second UWB communicating device are configured to be positioned at a plurality of spaced apart locations on a user, including at a first body location and a second body location;
- and
- a local controller placed in proximity to the plurality of UWB communicating devices, the local controller as a listening device is configured to (i) passively collect the localization packets sequentially transmitted by the respective UWB communicating devices and (ii) determine pair-wise distances between each respective pairs formed of the plurality of UWB communicating devices using the passively collected localization packets,
- wherein the collected localization packets including the pair-wise distances are used in a trained machine learning classifier model to determine a gesture classification of the user.
1 more element
A computer readable medium having instructions stored thereon for the classification of body gestures of a user, wherein the instructions, when executed by a processor, cause the processor to:
- receive, on a local controller placed in proximity to a plurality of ultrawide-band (UWB) communicating devices, packets of data from each of the plurality of UWB communication devices arranged as a plurality of nodes in a mesh configuration, wherein the plurality of UWB communicating devices includes a first UWB communicating device configured to sequentially transmit in each of a plurality of messaging cycles, a first localization packet of data to each of the other plurality of UWB communicating devices, including a second UWB communicating device, wherein the first UWB communicating device and the second UWB communicating device are configured to determine time associated values corresponding to receipts of the transmitted localization packets sent from the other UWB communicating devices, wherein the determined time associated value is included in the first localization packet, wherein the first UWB communicating device and second UWB communicating device are configured to be positioned at a plurality of spaced apart locations on a user, including at a first body location and a second body location;
- and
- determine pair-wise distances between each respective pairs formed of the plurality of UWB communicating devices using the collected localization packets, wherein the collected localization packets including the pair-wise distances are used in a trained machine learning classifier model to determine a gesture classification of the user.
A method for classifying body gestures of a user, the method comprising:
- receiving, on a local controller placed in proximity to a plurality of ultrawide-band (UWB) communicating devices, packets of data from each of the plurality of UWB communication devices arranged as a plurality of nodes in a mesh configuration, wherein the plurality of UWB communicating devices includes a first UWB communicating device that sequentially transmits in each of a plurality of messaging cycles, a first localization packet of data to each of the other plurality of UWB communicating devices, including a second UWB communicating device, wherein the first UWB communicating device and the second UWB communicating device determine time associated values corresponding to receipts of the transmitted localization packets sent from the other UWB communicating devices, wherein the determined time associated value is included in the first localization packet, wherein the first UWB communicating device and second UWB communicating device are positioned at a plurality of spaced apart locations on a user, including at a first body location and a second body location;
- and
- determining pair-wise distances between each respective pairs formed of the plurality of UWB communicating devices using the passively collected localization packets, wherein the collected localization packets including the pair-wise distances are used in a trained machine learning classifier model to determine a gesture classification of the user.
Dependencies are inferred from the claim text.