METHOD TO RECOGNIZE A GESTURE AND CORRESPONDING DEVICE

• US 20170315717A1
• Filed: 04/27/2016
• Published: 11/02/2017
• Est. Priority Date: 04/27/2016
• Status: Active Grant

• Current Assignee: Mogees Limited
• Original Assignee: Mogees Limited
• Inventors: Zamborlin, Bruno, Caramiaux, Baptiste, Cella, Carmine Emanuele
• Granted Patent: US 10,817,798 B2
• Time in Patent Office: Days
• 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 3/0416 Control or interface arrangements specially adapted for digitisers
• CPC Class Codes: G06F 3/0433 in which the acoustic waves are either generated by a movable member and propagated within a surface layer or propagated within a surface layer and captured by a movable member
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

The present disclosure concerns a method to recognize a gesture, namely a physical action, made in direct or indirect contact on a physical object, wherein said method comprises:

associating a device using said method in contact with said physical object;

supplying an entrance signal generated by a gesture made in direct or indirect contact on said physical object;

segmenting said entrance signal to obtain a group of signal segments and at least one physical value processed from said entrance signal;

processing at least a characteristic vector of signal features for each segment of said group of signal segments using said at least one physical value in order to learn metric function between gestures;

supplying a specific exit signal correlated to the specific gesture.

Further, the present invention concerns also a device to recognize a gesture using said method to recognize a gesture.

Claims

1. Method to recognize a gesture made in direct or indirect contact on a physical object, wherein said method comprises the following steps:

associating a device using said method in contact with said physical object;

supplying an entrance signal generated by a gesture made in direct or indirect contact on said physical object;

segmenting said entrance signal to obtain a group of signal segments and at least one physical value processed from said entrance signal;

processing at least a characteristic vector of signal features for each segment of said group of signal segments using said at least one physical value in order to learn metric function between gestures;

supplying a specific exit signal correlated to the specific gesture.

2. Method as in claim 1, further comprising a step of memorizing said characteristic vectors processed with an identification label of the gesture made on said physical object.

3. Method as in claim 1, further comprising a step of comparing a new gesture made, in direct or indirect contact with said physical object, with the gestures previously memorized to recognize if the new gesture made is one of the memorized gestures, said comparison step being made by evaluating and learning the metric function between said characteristic vectors, processed from the new gesture made, and said characteristic vectors processed from the gestures previously memorized.

4. Method as in claim 1, further comprising calculating for each element of said characteristic vector a metric function in order to quantify how much two gestures, identified by said identification labels, are similar, following the sub-steps:

ordering the characteristic vectors of the gestures according to their identification label;

applying the metric function to compare characteristic vectors of the gestures for classification;

assessing the metric function on the ordered characteristic vector of the gesture;

adjusting the metric function to minimize classification errors and iterate until the errors are minimum.

5. Method as in claim 1, further comprising a quality features step of said entrance signal, which can be carried out in parallel with the segmentation step, in order to obtain one or more adjustable descriptors of said gesture to assess the characteristics of said entrance signal.

6. Method as in claim 1, further comprising a sub-step (a) of extracting an envelope from said entrance signal to obtain a signal defining the functional form of the amplitudes of said entrance signal.

7. Method as in claim 6, further comprising a sub-step (b) of applying a low pass filter to said envelope of said entrance signal to obtain a signal with a functional profile of said envelope without noise coming from other sources or external agents that are not generated by or correlated to the gesture made on said physical object.

8. Method as in claim 6, further comprising a sub-step of selection and division of said filtered envelope of said entrance signal in said group of signal segments comprising segments or parts of said entrance signal, said segments being temporally detected if they are greater than or equal to a threshold value of amplitude.

9. Method as in claim 1, further comprising a sub-step of calculating said at least one physical value from said entrance signal, at least one of said physical value being the gesture group velocity of said entrance signal computed as the maximum in amplitude of the segmented signal.

10. Method as in claim 1, further comprising a sub-step of calculating the fast Fourier transform of said entrance signal to obtain a vector of spectral amplitudes.

11. Method as in claim 10, further comprising a sub-step of calculating one or more descriptors computed as the first three spectral moments from said vector of spectral amplitudes.

12. Method as in claim 1, wherein said exit signal is a digital, electric, computer or optical signal that commands and/or controls a device connected to said device using said method.

13. Method as in claim 1, wherein said exit signal contains the values of likelihoods of a gesture made on said physical object with the previously memorized gestures, said likelihoods gestures being calculated by comparing said characteristic vectors of the memorized gestures with the gesture made.

14. Device to recognize a gesture made in direct or indirect contact on a physical object, comprising:

a transducer associable in contact with said physical object and configured to detect the vibrations and/or sounds generated by the gesture made on said physical object;

a reception unit configured to receive an entrance signal from said transducer generated from said gesture made on said physical object;

a segmentation unit configured to receive said entrance signal and to obtain a group of signal segments and at least one physical value processed from said entrance signal;

a processing unit configured to calculate at least one characteristic vector of signal features for each segment of said group of signal segments and said at least one physical value;

a command unit configured to supply to a device a specific exit signal correlated to the specific gesture made.

15. Device as in claim 14, wherein said transducer associable in contact with said physical object is an adhesive piezo-transducer contact microphone configured to capture the sounds and/or the vibrations generated by a gesture made on said physical object.

16. Device as in claim 14, further comprising a memory unit configured to memorize said at least one characteristic vector, processed by said entrance signal, and an identification label defining the gesture made on said physical object.

17. Device as in claim 14, further comprising a comparison-learning unit configured to recognize if a new gesture made on said physical object is one of the gestures memorized in said memory unit.

18. Device as in claim 14, further comprising a quality features unit configured to obtain one or more adjustable descriptors of the gesture to assess the characteristics of said entrance signal supplied to it by said reception unit.