Alarm method and system based on voice events, and building method on behavior trajectory thereof

• US 8,237,571 B2
• Filed: 02/06/2009
• Issued: 08/07/2012
• Est. Priority Date: 11/26/2008
• Status: Active Grant

• Current Assignee: Industrial Technology Research Institute
• Original Assignee: Industrial Technology Research Institute
• Inventors: Wang, Yuh-Ching, Chiu, Yu-Hsien, Yan, Gwo Lang
• Primary Examiner(s): Lee, Benjamin C
• Assistant Examiner(s): PHAM, QUANG
• Application Number:
  US12/366,650
• Publication Number:
  US 20100127878A1
• Time in Patent Office: 1,278 Days
• Field of Search: 704/246, 704/270, 704/E21.001, 340/540, 340/506, 340/573.1, 348/143, 705/2
• US Class Current: 340/573.1
• CPC Class Codes: G08B 13/1672 using sonic detecting means, e.g. a microphone operating in the audio frequency range
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices

Abstract

Disclosed are an alarm method and system based on voice events, and a building method on behavior trajectory thereof. The system comprises a signal sensor, a voice-event detector and notice and alarm element. In the method, voice signals are captured from a remote unit in an environment. The captured voice signals are classified into at least a voice event. As such, an emergent-event notice is automatically transmitted out if one of predefined emergent events is detected. In the building method on behavior trajectory, messages on voice events are continuously recorded. When the number of the recorded voice events reaches a threshold, a behavior trajectory is constructed, in which a behavior consists of two or more voice events or a single voice event.

Claims

1. An alarm method based on voice events, said method comprising:

capturing a plurality of voice signals and sensing other sensed signals of an environment where a remote unit is located;

integrating the captured voice signals and the other sensed signals into at least a voice event by using a voice event detection method;

continuously detecting voice events through said voice event detection method and continuously recording the detected voice events;

generating a behavior trajectory through a behavior trajectory building method after accumulating a sufficient number of the detected voice events;

detecting a latest behavior through latest continuous voice events and comparing said latest behavior with a history of said behavior trajectory; and

issuing a notification alarm message when an abnormal behavior is detected;

wherein said behavior trajectory is generated by said behavior trajectory building method based on classifying qualified pre-selected strings of letters into a plurality of behavior types for building said behavior trajectory, said qualified pre-selected strings of letters being determined by naming each of the voice events with a letter to form letter sequences for the continuously detected voice events, finding identical strings in the letter sequences by comparing the letter sequences in a pair-wise manner, and qualifying each identical string of each type according to presence probability and length of the identical string.

2. The method as claimed in claim 1, wherein said voice event detection method at least includes:

performing voice endpoint detection on said captured voice signals to obtain a plurality of voice segments;

capturing at least a characteristic parameter of each of said plurality of voice segments; and

comparing said at least a characteristic parameter with pre-stored characteristic parameters for voice signal comparison to determine a corresponding voice event type.

3. The method as claimed in claim 1, wherein said method further includes an abnormal behavior decision method to continuously determine whether said latest behavior is an abnormal behavior or not, wherein a behavior consists of a single voice event, or two or more voice events.

4. The method as claimed in claim 1, wherein said behavior trajectory building method at least includes:

naming said detected voice events with letters to obtain a plurality of letter sequences representing said detected voice events;

comparing pair-wise said plurality of letter sequences to find at least an identical string;

computing presence probability and length of said at least an identical string and setting thresholds for said probability and said length to obtain at least a qualified pre-selected string;

classifying said at least a pre-selected string into a plurality of types, with each type representing a behavior type, by using a classification algorithm; and

inducing said at least a qualified pre-selected string of each type to build a behavior trajectory.

5. The method as claimed in claim 1, wherein said voice event detection method further includes detection of time and location of said at least a voice event and recording the time and location of said at least a voice event.

6. The method as claimed in claim 5, wherein the location of said at least a voice event is within a closed living space.

7. The method as claimed in claim 1, wherein a voice able to trigger an emergency event is a human meaningless involuntary voice or a meaningful voice indicating a human bumping into wall, floor or other objects.

8. A method for building a behavior trajectory, comprising:

continuously detecting voice events by integrating captured voice signals and other sensed signals using a voice event detection method and continuously recording the detected voice events;

naming the detected voice events with letters to obtain a plurality of letter sequences representing the detected voice events;

comparing pair-wise said plurality of letter sequences to find at least one identical string;

computing presence probability and length of said at last one identical string and setting thresholds for the probability and length to obtain at least a qualified pre-selected string;

classifying said at least a pre-selected string into a plurality of types, with each type representing a behavior type, by using a classification algorithm; and

inducing said at least a qualified pre-selected string of each type to build a behavior trajectory;

wherein detecting a latest behavior through latest continuous voice events and comparing said latest behavior with a history of said behavior trajectory; and

issuing a notification alarm message when an abnormal behavior is detected.

9. The method as claimed in claim 8, wherein said voice event detection method classifies a plurality of captured voice signals into at least one voice event.

10. An alarm system based on voice events, said system comprising:

a signal sensor for capturing one or more voice signals and sensing other sensed signals externally inputted, outputting the captured voice signals and the other sensed signals and recording locations of the captured voice signals and the other sensed signals;

a voice event detector connected to said signal sensor, for determining a plurality of corresponding voice events based on said locations and at least a characteristic parameter of the captured voice signals and the other sensed signals; and

a notice and alarm element connected to said voice event detector, for executing notification and alarming functions based on a behavior trajectory of said plurality of corresponding voice events;

wherein said system includes a behavior trajectory building module for generating said behavior trajectory through a behavior trajectory building method after accumulating a sufficient number of the voice events based on classifying qualified pre-selected strings of letters into a plurality of behavior types for building said behavior trajectory, said qualified pre-selected strings of letters being determined by naming each of the voice events with a letter to form letter sequences for continuously detected voice events, finding identical strings in the letter sequences by comparing the letter sequences in a pair-wise manner, and qualifying each identical string of each type according to presence probability and length of the identical string.

11. The system as claimed in claim 10, wherein a behavior consists of a single voice event or two or more voice events.

12. The system as claimed in claim 10, wherein said system further includes an abnormal behavior decision module for continuously detecting a latest behavior, comparing against a history of said behavior trajectory, and issuing an abnormal behavior alarm when detecting an abnormal behavior.

13. The system as claimed in claim 10, wherein said system is integrated with a toilet, and said signal sensor further captures a pressure signal and a switch signal triggered by a flushing switch or seat of said toilet.

14. The system as claimed in claim 10, wherein said system is integrated with a bed mattress, and said signal sensor further captures a pressure signal, a switch signal, a temperature signal humidity signal or a physiological signal triggered by a human or objects on said bed mattress.

15. The system as claimed in claim 10, wherein said system is integrated with a ceiling, and said signal sensor further captures at least an ultrasonic distance-measuring signal from an ultrasonic distance-measuring device installed on said ceiling.

16. The system as claimed in claim 10, wherein said system is integrated with a faucet, and said signal sensor further captures a pressure signal or a switch signal triggered by a switch of said faucet.

17. The system as claimed in claim 10, wherein said system is integrated with a wheelchair, and said signal sensor further captures a pressure signal or a switch signal triggered by a cushion, wheel or brake switch of said wheelchair.

18. The system as claimed in claim 10, wherein said system is integrated with a door or a window, and said signal sensor further captures a pressure signal or a switch signal triggered by opening and closing said door or said window.

19. The system as claimed in claim 10, wherein a voice able to trigger an emergency event is a human meaningless involuntary voice or a meaningful voice indicating a human bumping into wall, floor or other objects.

20. The system as claimed in claim 10, wherein said voice event is an ordinary voice event or an emergency event.

21. The system as claimed in claim 10, wherein said system is installed in a living space chosen from a group consisting of bathroom, hospital room, household bedroom and living room.

22. The system as claimed in claim 20, wherein said emergency event is a pre-defined event selected from the events triggered by the voices of falling during getting on or off a bed, falling during getting on or off a wheelchair, being tripped or slipping when walking, choking when drinking, temporarily stopping breathing, fighting or having an argument, crying or moaning, calling or calling for help, sudden illness attack, wrecking, breaking or falling objects, being unable to finish the use of a toilet, and entering a dangerous zone.

23. The system as claimed in claim 20, wherein said ordinary voice event is an event selected from the events triggered by the voices of walking or movement, getting on or off bed, getting on or off a wheelchair, opening or closing a door, using a faucet, using a toilet, flushing a toilet, and drinking water.