DEVICE AND METHOD FOR RECOGNIZING GESTURES FOR A USER CONTROL INTERFACE

• US 20150370332A1
• Filed: 12/09/2013
• Published: 12/24/2015
• Est. Priority Date: 12/12/2012
• Status: Active Grant

• Current Assignee: Sagemcom Broadband SAS (Sagemcom Holding SAS)
• Original Assignee: Sagemcom Broadband SAS (Sagemcom Holding SAS)
• Inventors: EVRARD, Etienne, COLINET, Frdric
• Granted Patent: US 10,802,593 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 3/017 Gesture based interaction, e.g. based on a set of recognized hand gestures interaction based on gestures traced on a digitiser G06F3/04883
• CPC Class Codes: G06F 3/0346 with detection of the device orientation or free movement in a 3D space, e.g. 3D mice, 6-DOF [six degrees of freedom] pointers using gyroscopes, accelerometers or tilt-sensors
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

In the context of a user interface control, a gesture-recognition device: receives gyroscopic data representing a gesture executed with a dedicated instrument including a gyroscopic sensor; determines a correlation between the received gyroscopic data and gyroscopic data relating to a supervised learning and pre-recorded into a database; recognizes or not the executed gesture according to said correlation, the only data representing the executed gesture taken into account being said gyroscopic data; transposes each recognized gesture into a user interface command.

Claims

1. A gesture-recognition method for recognizing gestures in the context of user interface control, said method being implemented by a gesture-recognition device, the method comprising receiving gyroscopic data representing a gesture executed with a dedicated instrument comprising a gyroscopic sensor, wherein the gesture recognition device includes a network of artificial neurons comprising output neurons associated respectively with the gestures that said device is adapted for recognizing, and wherein the method further comprises:

determining, by the network of artificial neurons, a correlation between the received gyroscopic data and gyroscopic data relating to a supervised learning and pre-recorded in a database, the network of artificial neurons implementing an activation function of the sigmoid type and, for the received gyroscopic data and for each output neuron, supplying a probability that said received gyroscopic data correspond to the gesture associated with said output neuron;

recognizing or not the executed gesture according to said correlation, the only data representing the executed gesture taken into account being said gyroscopic data; and

transposing each recognized gesture into a user interface command.

2. The gesture recognition method according to claim 1, wherein, the received gyroscopic data representing measurements made by the gyroscopic sensor between two actions detected on a button of said dedicated instrument, receiving the gyroscopic-data is followed by deleting a predefined number of gyroscopic data from the start of the measurements and/or a predefined number of gyroscopic data to the end of the measurements.

3. The gesture-recognition method according to claim 1, wherein it comprises a sampling operation by equally distributed deletion or equally distributed preservation of gyroscopic data among the received gyroscopic data.

4. The gesture-recognition method according to claim 1, wherein the gyroscopic data relating to the supervised learning and pre-recorded in the database correspond to all the following gestures:

a horizontal gesture to the right (401);

a horizontal gesture to the left (402);

a vertical gesture upwards (403); and

a vertical gesture downwards (404).

5. The gesture-recognition method according to claim 4, wherein, the gyroscopic data received representing measurements made by the gyroscopic sensor on three axes of an inertial reference frame, said gesture-recognition method comprises normalizing the received gyroscopic data so that:

for each original gyroscopic data item defined with respect to a given axis, the normalized gyroscopic data item is equal to said original gyroscopic data item minus the minimum value of all the original gyroscopic data defined with respect to said given axis, the whole divided by the minimum value of all the original gyroscopic data defined with respect to said given axis minus the minimum value of all the original gyroscopic data defined with respect to said given axis.

6. The gesture-recognition method according to claim 4, wherein set of gestures further comprises:

a gesture corresponding to the reproduction of the letter “

V”

; and

a circular gesture starting from a high position and beginning the rotation towards the left.

7. The gesture-recognition method according to claim 6, wherein, the received gyroscopic data representing measurements made by the gyroscopic sensor on three axes of an inertial reference frame, said gesture-recognition method comprises normalizing the received gyroscopic received so that:

for each original gyroscopic data item defined with respect to a given axis, the normalized gyroscopic data item is equal to said original gyroscopic data item minus the minimum value of all the original gyroscopic data defined with respect to said given axis, the whole divided by the minimum value of all the original gyroscopic data defined with respect to said given axis minus the mean of all the original gyroscopic data defined with respect to said given axis, the whole divided by the standard deviation of all the original gyroscopic data defined with respect to said given axis.

8. A gesture-recognition device for receiving gyroscopic data representing a gesture executed with a dedicated instrument comprising a gyroscopic sensor, wherein said gesture-recognition device comprises a network of artificial neurons comprising output neurons associated respectively with the gestures that said device is adapted for recognizing and said gesture-recognition device is further configured for:

thanks to the network of artificial neurons, determining a correlation between the gyroscopic data received and gyroscopic data relating to a supervised learning and pre-recorded in a database, the network of artificial neurons being adapted for implementing an activation function of the sigmoid type and, for the gyroscopic data received and for each output neuron, supplying a probability that said received gyroscopic data correspond to the gesture associated with said output neuron;

recognizing or not the executed gesture according to said correlation, the only data representing the executed gesture taken into account being said gyroscopic data; and

transposing each recognized gesture into a user interface command.