Method and system to count movements of persons from vibrations in a floor

• US 10,339,438 B2
• Filed: 07/26/2018
• Issued: 07/02/2019
• Est. Priority Date: 01/11/2017
• Status: Active Grant

• Current Assignee: Thomas Danaher Harvey
• Original Assignee: Thomas Danaher Harvey
• Inventors: Harvey, Thomas Danaher
• Primary Examiner(s): King, Curtis J
• Application Number:
  US16/046,136
• Publication Number:
  US 20180357525A1
• Time in Patent Office: 341 Days
• Field of Search: None
• CPC Class Codes: F41A 17/06 Electric or electromechanical safeties F41A17/04, F41A17/08 take precedence
• CPC Class Codes: F41A 17/063 comprising a transponder
• CPC Class Codes: F41A 17/066 having means for recognizing biometric parameters, e.g. voice control, finger print or palm print control
• CPC Class Codes: G06K 17/0022 arrangements or provisious for transferring data to distant stations, e.g. from a sensing device "transfer between computer elements G06F13/00 "; data-transmission H04L
• CPC Class Codes: G06K 19/0711 the source being mechanical or acoustical
• CPC Class Codes: G06K 19/14 the marking being sensed by radiation
• CPC Class Codes: G06K 7/10297 arrangements for handling protocols designed for non-contact record carriers such as RFIDs NFCs, e.g. ISO/IEC 14443 and 18092 cryptographic protocols H04L9/00; network security protocols H04L63/00; real-time communication protocols in data switching networks H04L65/00; network protocols for data switching network services H04L67/00
• CPC Class Codes: G06M 1/27 for representing the result of count in the form of electric signals, e.g. by sensing markings on the counter drum
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G06T 1/0007 Image acquisition
• CPC Class Codes: G06T 2207/30236 Traffic on road, railway or crossing
• CPC Class Codes: G06T 2207/30242 Counting objects in image
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
• CPC Class Codes: G06V 40/25 Recognition of walking or running movements, e.g. gait recognition

Abstract

A system and method for counting persons using passages entering or exiting an area by analyzing vibrations in the floor with sensors and a machine learning system. The machine learning system uses a model, usually implemented as a neural network on a processor. The network is trained in levels and implemented in layers. Different levels classify and analyze vibrations by timing and frequency, by movements of persons, and by identity of persons The same person is identified by patterns in the vibrations and the vibrations are correlated to determine when a person uses a combination of passages and is thereby counted. Location information for the person is used to identify persons in places and doing activities of interest. The model may be trained on one processor and then downloaded to another processor for evaluation. Additional sensors and levels of training may be implemented on the latter processor.

Claims

1. A method of counting persons comprising:

(a) producing signals from a sensor adapted to detect vibrations made by persons crossing a floor under an area connecting at least three passageways;

(b) evaluating the signals with a machine learning system trained to detect patterns of movement of persons between the passageways from vibrations in the floor;

(c) determining from outputs of the machine learning system produced on the basis of the signals that a person has entered the area by an identified passageway and that the person has exited the area by an identified passageway; and

(d) producing a count of the number of persons that have passed through the area using the combination of passageways identified by the outputs.

2. The method of claim 1 wherein:

the count is restricted to persons who have been identified on the basis of signals from the sensors as being in a portion of the floor during a time range.

3. The method of claim 2 wherein:

the time range is determined by an event other than vibrations in the floor.

4. The method of claim 1 wherein:

the machine learning system is trained to detect locations of sources of signals from vibrations produced by movements of persons crossing a floor to identify signals produced at different times by a particular person as being produced by the same person.

5. The method of claim 1 wherein:

the machine learning system is implemented on a first processor to train the system to identify movements from signals from a sensor to detect vibration and

a second processor to identify persons producing vibrations at different times as being the same person subsequent to downloading a model comprised in the machine learning system from the first processor to the second processor.

6. The method of claim 1 wherein:

the first processor uses data from a sensor to detect vibrations in a floor to train a model comprised in the machine learning system and the second processor uses data from sensor of claim 1 to evaluate the model.

7. A system for counting persons comprising:

(a) A sensor to produce signals from vibrations,

(b) a floor connecting a set of at least one area by which a person may at least one of enter the floor and exit the floor;

(c) a processor to evaluate a machine learning model wherein;

the model comprises a first layer wherein outputs of the layer are generated from an evaluation of the sensor signals on the basis of training to classify events based on timing and frequency determinations,

the model comprises a layer to evaluate outputs of the first layer to produce an output identifying a multiplicity of locations of origin of signals produced by a particular person on the basis of training from signals collected from the sensor, and

the processor produces an output of the model on the basis of the locations which signifies that a person has entered using a particular one of the areas of the set and the person has exited using a particular one of the areas of the set; and

(d) an output device for at least one of signaling a count produced on the basis of the output of persons which have entered by a particular one of the areas of the set and exited by a particular one of the areas of the set and of displaying a count of persons which have entered by a particular one of the areas of the set and exited by a particular one of the areas of the set.

8. The method of claim 7 wherein:

the count is restricted to persons who have been identified on the basis of signals from the sensors as being in a portion of the floor during a time range.

9. The method of claim 8 wherein:

the time range is determined by an event other than vibrations in the floor.

10. The system of claim 7 wherein:

the second layer of the model comprises two sub-layers with one sub-layer of the second layer receiving outputs signifying locations of persons from a second sub-layer of the second layer of the model and producing outputs signifying that the same person has produced vibrations at a multiplicity of the signified locations.

11. The system of claim 7 wherein:

a second processor trains the model,

the model is downloaded to said first processor, and said first processor evaluates inputs from the sensor and produces the output of the model.

12. The system of claim 11 wherein:

said first processor trains the model subsequent to the training of the second processor.

13. The system of claim 7 wherein:

a second sensor to detect vibrations provides data used to train the model.

14. A system for counting persons comprising:

(a) a sensor to produce signals from vibrations,

(b) a floor connecting a set of at least two areas each of which a person may use to at least one of enter the floor and exit the floor;

(c) a processor to evaluate a machine learning model wherein;

the model comprises a first layer wherein outputs of the layer are generated from an evaluation of the sensor signals on the basis of training to classify events based on timing and frequency determinations,

the model comprises a second layer to evaluate outputs of the first layer to produce an output identifying a multiplicity of locations of origin of signals produced by a particular person on the basis of training from signals collected from the sensor, and

the processor produces an output of the model on the basis of the locations which signifies that a person has entered using a particular one of the areas of the set and the person has exited using a particular one of the areas of the set; and

(d) an output device for at least one of signaling a count of a set of persons produced on the basis of the output wherein the set of persons consists of persons who have entered by a particular one of the areas of the set of areas and exited by a particular one of the areas of the set of areas and of displaying a count of persons which have entered by a particular one of the areas of the set of areas and exited by a particular one of the areas of the set of areas, and wherein the set of persons is limited to persons who have been identified as having been in a specific area of the floor in a specific time range.

15. The system of claim 14 wherein:

the time range is determined by an event other than vibrations sensed in the floor.

16. The system of claim 14 wherein:

the second layer of the model comprises two sub-layers with one sub-layer of the second layer receiving outputs signifying locations of persons from a second sub-layer of the second layer of the model and producing outputs signifying that the same person has produced vibrations at a multiplicity of the signified locations.

17. The system of claim 14 wherein:

a second processor trains the model,

the model is downloaded to said first processor, and

said first processor evaluates inputs from the sensor and produces the output of the model.

18. The system of claim 17 wherein:

said first processor trains the model subsequent to the training of the second processor.

19. The system of claim 14 wherein:

a second sensor to detect vibrations in the floor provides data used to train the model.