Power-efficient hand tracking with time-of-flight sensor

A method of operating a time-of-flight (TOF) sensor, the method comprising:

• operating the TOF sensor in a low power mode by repeatedly performing a low power mode sequence comprising a depth frame and an amplitude frame;
• estimating three-dimensional (3D) positions of keypoints of a target object while operating in the low power mode, wherein estimating the 3D positions comprises computing two-dimensional (2D) positions of the keypoints based on an amplitude map captured during the amplitude frame and estimating the 3D positions based on the 2D positions and a previously captured depth map;
• determining that an activation condition is satisfied, wherein the activation condition comprises a likelihood that the target object is interacting in a Z dimension;
• switching from the low power mode to a high accuracy mode in response to determining that the activation condition is satisfied, wherein operating in the high accuracy mode includes repeatedly performing a high accuracy mode sequence comprising multiple depth frames;
• and
• computing the 3D positions of the keypoints while operating in the high accuracy mode.

A method of operating a time-of-flight (TOF) sensor, the method comprising:

• operating the TOF sensor in a low power mode by repeatedly performing a low power mode sequence comprising a depth frame and an amplitude frame;
• computing a depth map during the depth frame by capturing data during at least one illuminated subframe and an intensity subframe in which an illumination unit is disabled, and subtracting ambient information captured during the intensity subframe from the data captured during the at least one illuminated subframe to reduce noise;
• estimating three-dimensional (3D) positions of keypoints of a target object based on the amplitude frame and the depth map while operating in the low power mode;
• determining that an activation condition is satisfied, wherein the activation condition comprises a likelihood that the target object is interacting in a Z dimension;
• switching from the low power mode to a high accuracy mode in response to determining that the activation condition is satisfied, wherein operating in the high accuracy mode includes repeatedly performing a high accuracy mode sequence comprising multiple depth frames;
• and
• computing the 3D positions of the keypoints while operating in the high accuracy mode.