Facebook Patent | Armband for Tracking Hand Motion Using Electrical Impedance Measurement

Patent: Armband for Tracking Hand Motion Using Electrical Impedance Measurement

Publication Number: 20190033974

Publication Date: 2019-01-31

Applicants: Facebook

Abstract

A system includes a wearable device including sensors arranged at different locations on the wearable device. Each sensor measures electrical signals transmitted from a wrist or arm of a user. A position computation circuit is coupled to the sensors. The position computation circuit computes, using information derived from the electrical signals with a machine learning model, an output that describes a hand position of a hand of the wrist or arm of the user.

Background

Bodily motion such as gestures can be an appealing way for a user to interact with computing devices, since gestures are already a natural part of how people communicate. A hand gesture tracking device may enable a user to control or interact with computing devices without physically touching them. For example, a gesture tracking device may be used by a user to control memory or display devices in a local or remote environment.

Gesture recognition systems can be unwieldy, constraining, and uncomfortable to use. For example, glove-like devices that use mechanical signals to track hand motion may disturb the haptic sense of the hand. They may also be difficult to customize to different users having varying hand size. Camera-based gesture tracking systems detect the user’s hand area from captured images, but often require the user’s entire hand to be in the camera view range. They may also suffer failure when there is object occlusion in an image.

Gesture trackers are thus awkward to use, difficult to manage, and subject to distortion from the system arrangement or surrounding environment. The cost of traditional gesture trackers may also be high, thus limiting user adoption rate. Furthermore, gesture recognition systems are limited to a small set of predefined gestures. For these and other reasons, such gesture recognition systems are unsuitable for effective hand tracking for head-mounted displays (HMDs).

Summary

Embodiments relate to a wearable device for tracking human hand motion using electrical impedance measurement of the user’s arm. Sensors located on the wearable device receive electrical signals from the user’s wrist or arm corresponding to a hand position of the user. Information is extracted from the electrical signals (e.g., impedance values indicating the state of the user’s arm structures such as tendons, and thus also indicating the hand position or pose of the user). The impedance or other extracted values from the electrical signals may be used as input to a position computation circuit implementing a machine learning model. The position computation circuit determines the hand position of the user from the inputs using the machine learning model. In some embodiments, the wearable device includes an inertial measurement unit (IMU) that measures motion of the user’s arm, and provides inertial signals to the position computation circuit as an input for determination of the hand position. In some embodiments, the machine learning model is trained using ground truth hand positions extracted from video images of the user’s hand taken concurrently with the capture of the electronic signals.

In one embodiment, a system includes a wearable device including sensors arranged on the wearable device. Each sensor measures electrical signals transmitted from a wrist or arm of a user. A position computation circuit is coupled to the sensors. The position computation circuit computes, using information derived from the electrical signals with a machine learning model, an output that describes a hand position of a hand of the wrist or arm of the user.

In one embodiment, the position computation circuit is located in the wearable device.

In one embodiment, the position computation circuit is located in a host system or other computing device external to the wearable device.

In one embodiment, each sensor includes an electrode and a conductive agent located between the electrode and the wrist or arm of the user. Each sensor can transmit an alternative current (AC) signal, a direct current (DC) signal, or a wide-bandwidth AC signal including multiple frequencies.

In one embodiment, the sensors are arranged in a grid array or a checkerboard pattern.

In one embodiment, the information derived from the electrical signals includes aggregate values based on an electrical impedance measured between each pair of sensors. The electrical impedance is determined based on probe signals transmitted into the user’s wrist or arm by the pair of sensors and the measured electrical signals. In one embodiment, the derived information may include a shape of a wave of the electrical signals, a frequency-domain representation of the electrical signals, or a time-domain sample of the electrical signals.

In one embodiment, the position computation circuit extracts features from the information derived from the electrical signals. The features include one or more of angles between joints defining the hand position of a hand of the wrist or arm of the user, and a reduced representation of a change in the hand position of a hand of the wrist or arm of the user. The reduced representation defines a difference between a present hand position of the hand of the wrist or arm of the user from a previous hand position of the hand of the wrist or arm of the user.

In one embodiment, each sensor further transmits a probe signal into the user’s wrist or arm by varying a time period of the probe signal, an amplitude of the probe signal, or a phase of the probe signal.

In one embodiment, a first sensor transmits a probe signal into the user’s wrist or arm by staggering the transmission of the probe signal with respect to transmission of other probe signals by other sensors.

In one embodiment, an inertial measurement unit generates inertial signals corresponding to movement of the wearable device, and the user’s arm. The position computation circuit further computes, using information derived from the inertial signals with the machine learning model, the output that describes the hand position of the hand of the wrist or arm of the user.

In one embodiment, the inertial measurement unit includes one or more of a gyroscope, an accelerometer, and a magnetometer.

In one embodiment, one or more cameras generate image signals of the user’s hand, such as images that show the hand position from one or multiple angles. The position computation circuit further computes, using information derived from the image signals with the machine learning model, the output that describes the hand position of the hand of the wrist or arm of the user.

In one embodiment, the camera is a depth camera, a Red-Green-Blue (RGB) camera, an Infrared camera, or a camera mounted on an HMD.

In one embodiment, an inertial measurement unit generates inertial signals corresponding to movement of the wearable device. The position computation circuit further trains the machine learning model to generate, using information derived from the inertial signals, the output that describes the hand position of the hand of the wrist or arm of the user.

In one embodiment, a camera generates image signals of the hand of the wrist or arm of the user. The position computation circuit further trains the machine learning model to generate, using information derived from the image signals with the machine learning model, the output that describes the hand position of the hand of the wrist or arm of the user.

In one embodiment, the computed output further describes forces exerted by the hand of the user on objects touching the hand of the user. The structures of the user’s arm, such as the tendons, are in different states when the hand is exerting force compared with when the hand is relaxed even if the overall position of the hand looks similar. The use of electrical signals for hand tracking thus allows for even more types of data extraction than possible with other types of sensor tracking systems.

In one embodiment, the position computation circuit further receives image signals from a camera mounted on an HIVID. Comparison signals are determined by comparing the image signals with the computed output. The comparison signals are transmitted to a host system.

In one embodiment, the computed output includes parameters of a hand shape model. The parameters correspond to joints of the hand of the wrist or arm of the user, edges between pairs of the joints, angles between pairs of the edges, and a mesh including vertices and for each vertex, a distance between the vertex and one or more joints.

发表评论

电子邮件地址不会被公开。 必填项已用*标注