JP5626372B2 - Event detection system - Google Patents

Description

  The present invention relates to an event detection system that detects an event that has occurred.

  The need for technology related to crime prevention and watching is increasing year by year. Factors behind this are the aging of households, the increase in single-person households, and the increasing remoteness of households between generations as a result of urban concentration of employment. In the technology related to crime prevention and watching, when an event such as an abnormal situation occurs, a technology for detecting the fact is important.

  As such a technique, a technique for notifying that an abnormal event has been detected when an abnormal sound is detected is known (see, for example, Patent Document 1). A technique for detecting an event by sound, such as the technique described in Patent Document 1, has a high merit in terms of privacy and cost.

  For example, sound analysis is performed after irreversible conversion. Therefore, according to such a technique, it is not necessary to store sound signal information while maintaining an audible state. For this reason, depending on the technology for detecting an event by sound, the possibility of sound leaking to the outside is low, and the advantage in privacy is high.

  For example, a general-purpose microphone or a piezoelectric element can be used for the sensing device. Therefore, depending on the technology for detecting an event by sound, the cost can be significantly reduced as compared with a technology that uses a camera or a dedicated device as a sensing device.

  For example, the amount of sound data per second is about several tens of kilobytes. A technique for reducing the weight of sound processing has already been established, as represented by Fast Fourier Transform. Such a reduction in processing cost also contributes to a reduction in processor cost.

  For example, a sensing device such as a microphone can be installed relatively easily as compared with a camera installed at a high place or a dedicated device installed by an expert. For this reason, a financial burden and a physical burden can be reduced.

  Known techniques for detecting events by sound have problems with sound identification accuracy. More specifically, depending on a known technique, an intruder may make adjustments such as reducing the generated sound, thereby overlooking an event that should be detected.

  One method for solving such a problem is disclosed in Patent Document 2. The device disclosed in Patent Document 2 is a sound pressure sensor that detects a change in air flow accompanying opening / closing of a door as a low frequency sound instead of a door opening / closing sound, or a pneumatic sensor that detects a change in air pressure as vibration. Is used to detect the opening and closing of the door.

Japanese Unexamined Patent Publication No. 2002-31860 Japanese Unexamined Patent Application Publication No. 2009-093477

  The importance of an event varies depending on whether the event that has occurred is an event that has occurred due to a human action or an event that has occurred due to the movement of a non-human animal or a natural phenomenon. In other words, the subject that causes the occurrence of an event is an important factor in determining the meaning and importance of the event. However, the techniques described in Patent Document 1 and Patent Document 2 do not disclose a method for determining what is the subject of the occurrence of an event.

In order to solve the above-described problem , according to the first aspect of the present invention, there is provided an event detection system for detecting an event that has occurred, and extracting a time-series feature vector from signal information of acquired sound The sound identification unit for identifying the type of acquired sound by comparing the time series feature amount vector extracted by the unit, the feature amount extraction unit, and the model data for each type of sound, and received from the sound identification unit recorded in time series type of acquired sound and, based on the time series generation pattern of acquisition sound types, and human judgment unit judges human property, an output in accordance with the determination result of human determination unit determines And an output unit for performing.

  As is clear from the above description, depending on the event detection system of the present invention, after identifying the type of sound, based on the presence or occurrence pattern of sound caused by human action, which is included before and after the sound, It is possible to determine the human nature of the sound. Specific examples of actions include physical movements such as movement and applause, movements that do not involve movement such as vocalization and sneezing, and machine operations. However, the present invention is not limited to this. Therefore, depending on the event detection system of the present invention, for example, when human footsteps are detected before and after the sound of opening and closing the door, the door does not open and close due to natural phenomena such as wind, but artificially It is determined that this was done.

  Further, in the event detection system of the present invention, human artifacts are determined by sound. Therefore, some event detection systems of the present invention are advantageous in terms of privacy and cost.

It is a figure which shows an example of the utilization environment of the event detection apparatus which concerns on the 1st Embodiment of this invention. It is a figure which shows an example of the block configuration of the event detection apparatus shown in FIG. It is a figure which shows an example of the data stored in the artifact determination database shown in FIG. 2 in a table format. It is a figure which shows an example of the operation | movement flow of the event detection apparatus shown in FIG. It is a figure which shows an example of the utilization environment of the event detection apparatus which concerns on the 2nd Embodiment of this invention. It is a figure which shows an example of the block configuration of the event detection apparatus shown in FIG. It is a figure which shows an example of the data stored in the artifact determination database shown in FIG. 6 in a table format. It is a figure which shows an example of the operation | movement flow of the event detection apparatus shown in FIG. It is a figure which shows an example of the utilization environment of the event detection apparatus which concerns on the 3rd Embodiment of this invention. It is a figure which shows an example of the block configuration of the event detection apparatus shown in FIG. It is a figure which shows an example of the data stored in the artifact determination database shown in FIG. 10 in a table format. It is a figure which shows an example of the operation | movement flow of the event detection apparatus shown in FIG. It is a figure which shows an example of the block configuration of the event detection apparatus which concerns on the 4th Embodiment of this invention. It is a figure which shows an example of the block configuration of another example of the event detection apparatus of the 4th Embodiment of this invention. It is a figure which shows an example of the data stored in the artifact determination database shown in FIG. 14 in a table format. It is a figure which shows an example of the utilization environment of another example of the event detection apparatus of the 4th Embodiment of this invention. It is a figure which shows an example of the block configuration of the event detection apparatus shown in FIG. It is a figure which shows an example of the utilization environment of the event detection apparatus which concerns on the 5th Embodiment of this invention. It is a figure which shows an example of the data stored in the artifact determination database of the event detection apparatus shown in FIG. 18 in a table format. It is a figure which shows an example of the operation | movement sequence of a telephone set, the personal computer, a mobile telephone, and an event detection apparatus in the 5th Embodiment of this invention.

  Hereinafter, the present invention will be described through embodiments of the invention. However, the following embodiments do not limit the invention according to the scope of claims, and the features described in the embodiments are not limited thereto. Not all combinations are essential for the solution of the invention.

(First embodiment)
FIG. 1 shows an example of the usage environment of the event detection apparatus 100 according to the first embodiment of the present invention. The event detection device 100 detects an event that has occurred. More specifically, the event detection apparatus 100 is installed in the vicinity of the door 901. The event detection apparatus 100 collects opening / closing sounds of the door 901. In addition, the event detection apparatus 100 collects footstep sounds of human beings near the door 901. Further, the event detection device 100 may be an example of an “event detection system” in the embodiment of the present invention.

FIG. 2 shows an example of a block configuration of the event detection apparatus 100. The event detection apparatus 100 includes a sound input element 110, a feature amount extraction unit 120, a sound identification unit 130, an artifact determination unit 140, an output unit 150, a sound identification model database 191 and an artifact determination database 192.
The function and operation of each component will be described below.

  The sound input element 110 converts sound into signal information. Then, the sound input element 110 sends the signal information of the sound to the feature amount extraction unit 120. The sound signal information includes a continuous electric signal indicating a time-series change in sound pressure, and discrete sound data obtained by sampling and quantizing the electric signal. Specific examples of the sound input element 110 include a microphone and a piezoelectric element.

  When receiving the sound signal information from the sound input element 110, the feature amount extraction unit 120 extracts a feature amount determined to be necessary for sound identification from the sound signal information, and extracts a time-series feature amount vector. Is sent to the sound identification unit 130 as data indicating Specific examples of the feature amount include sound pressure average, sound pressure maximum value, sound pressure dispersion, number of zero point crossings, peak frequency, frequency spectrum, pitch, MFCC (Mel-Frequency Cepts Coefficients), MP (Matching Persitut), These primary differences, secondary differences, and the like. However, the present invention is not limited to this.

  When the sound identification unit 130 receives data indicating the time-series feature amount vector from the feature amount extraction unit 120, the time-series feature amount vector indicated by the data and the data stored in the sound identification model database 191. Is analyzed and the type of the sound is specified. Then, the sound identification unit 130 sends data indicating the type of the sound to the artificial determination unit 140. Specific identification methods include hidden Markov model (HMM: Hidden Markov Model), GMM (Gaussian Mixture Model), SVM (Support Vector Machine), likelihood determination based on statistical model such as Bayesian estimation, DP (Dynamic Prog) And similarity determination based on distance measurement between data such as DTW (Dynamic Time Warping). However, the present invention is not limited to this.

  Upon receiving the data indicating the sound type from the sound identification unit 130, the artifact determination unit 140 records the sound type indicated by the data in time series. Next, the human artifact determination unit 140 determines the human artifact by comparing and analyzing the time series occurrence pattern of the recorded sound type with the identification sound and the characteristic sound pattern stored in the human artifact determination database 192. . Then, the artifact determination unit 140 sends data indicating the determination result to the output unit 150. The artifact determination unit 140 includes a buffer for recording the type of sound generated during a predetermined time in order to determine the artifact using the time series generation pattern of the type of sound.

  When the output unit 150 receives data indicating the determination result from the artifact determination unit 140, the output unit 150 executes output according to the determination result indicated by the data. Specific examples include storage in a database, display, sound, identification sound presentation to the user through light, transmission of a mail to a specific mail address, transmission to a specific telephone number, transmission of a packet to an external device, etc. It is done.

  The sound identification model database 191 stores data serving as a model for comparing and analyzing sound feature vectors and specifying the sound type. Specific examples of model data include feature parameters, learned model parameters of a statistical model such as HMM, and the like. However, the present invention is not limited to this. In general, model data is prepared for each sound type to be identified, both for the feature amount and the learned model parameter.

  The artifact determination database 192 stores identification sounds, characteristic sound patterns, and artifacts in association with each other. The characteristic sound pattern indicates a time-sequential generation pattern of characteristic sounds for determining artifacts. Artificiality indicates the presence or absence of humanity in the generation of the identification sound when the occurrence of the characteristic sound pattern is detected.

  FIG. 3 shows an example of data stored in the artificial determination database 192 in a table format. In the identification sound, information “door open / close sound” is stored. In the characteristic sound pattern, information “footstep: front and back” is stored. In the artifact, information “present” is stored. These pieces of information mean that if “door open / close sound” is identified and “footstep sound before and after” is detected, it is determined that the door open / close signal, which is the identification sound, is “artificial”. ing.

  In the present embodiment, an example in which one data combination is registered in the artifact determination database 192 is shown, but the present invention is not limited to this. A combination of a plurality of patterns of data may be registered in the artifact determination database 192. In the present embodiment, as the characteristic sound pattern, the case where binary determination is performed on the condition of the presence or absence of detection of the characteristic sound has been described, but the present invention is not limited thereto. Likelihood determination based on a statistical model such as HMM, GMM, SVM, Bayesian estimation, or the like may be performed.

FIG. 4 shows an example of an operation flow of the event detection apparatus 100. The sound input element 110 included in the event detection apparatus 100 converts the air vibration of the detected sound into sound signal information (S101).
Subsequently, the feature quantity extraction unit 120 extracts a feature quantity from the signal information of the converted sound (S102). Thereafter, the sound identification unit 130 compares and analyzes the extracted feature amount and the model data registered in the sound identification model database 191 to identify the sound type (S103).
Subsequently, the human nature determination unit 140 records the identified sound types in time series, and generates the time series generation patterns of the sound types and the identification sounds and feature sound patterns registered in the human nature determination database 192. A comparative analysis is performed to determine the artifacts (S104). Finally, the output unit 150 performs output based on the identification sound and the presence or absence of artifacts (S105).

  With the above-described configuration and operation, the event detection apparatus 100 can determine the artifact of the identification sound while considering the privacy and suppressing the cost. That is, when the user opens / closes the door, footsteps are generated at a timing near the door opening / closing sound. On the other hand, when the door is opened / closed by wind or animal movement, footsteps are not generated at the timing near the door opening / closing sound. Therefore, it is possible to detect the door opening / closing performed artificially by using the event detection device 100 to recognize the difference in the generation pattern of the characteristic sound by using the door opening / closing sound as the identification sound and the footstep as the characteristic sound. . That is, it is possible to identify an artificial event without forcing the user to perform a special action.

(Second Embodiment)
FIG. 5 shows an example of the usage environment of the event detection apparatus 200 according to the second embodiment of the present invention. The event detection apparatus 200 can identify the person who has made a sound in addition to human artifacts. The event detection device 200 is installed at the entrance of the house and collects sounds generated at and near the entrance. In this embodiment, when the door 901 is opened and closed, a sound indicating the operation is emitted.
In addition, footsteps are emitted from users 902 in and around the room. In addition, when the door lock 903 is opened and closed using the key 904, an opening / closing lock sound is generated. Further, a bell 905 is associated with the key 904, and a bell sound is emitted from the bell 905 when the key 904 is used.
In other words, the event detection device 200 can detect “door open / close sound”, “human footsteps”, “open / close lock sound”, and “bell sound”, respectively. The event detection device 200 may be an example of an “event detection system” in the embodiment of the present invention.

  When the user 902 opens the door 901, the user 902 approaches the door 901 before opening the door 901, and moves away from the door 901 after opening the door 901. For this reason, the footstep sound of the user 902 is emitted before and / or after the door 901 is opened and closed. If the door 901 is locked, the key 904 has a bell 905. For this reason, a bell sound is emitted before and / or after the locking sound when the door 901 is opened and closed.

  FIG. 6 shows an example of a block configuration of the event detection apparatus 200. Among the constituent elements of the event detection apparatus 200, constituent elements having the same names and having the same reference numerals as the constituent elements of the event detection apparatus 100 show similar functions and operations.

  Differences between the event detection device 200 and the event detection device 100 will be described. In the artifact determination database 292 of the event detection apparatus 200, a person who is associated with the identification sound and the characteristic sound pattern is further associated and registered. Further, the human artifact determination unit 240 determines a person who has emitted the identification sound based on the characteristic sound generation pattern registered in the human artifact determination database 292.

  FIG. 7 shows an example of data stored in the artificial determination database 292 in a table format. In the example shown in FIG. 7, the artificiality determination information when “door opening / closing sound” and “opening / closing lock sound” are identified is registered. If “footsteps” are detected before and after the “door opening / closing sound” is identified, it is determined that the door opening and closing, which is the identification sound, is artificial, but the characteristic sound determines the person. It is stated that it can not be done. In addition, when “bell sound” is detected before and after the “opening and closing lock sound” is identified, the opening and closing lock that is the identification sound is artificial, and the opening and closing lock was further performed. It is described that it can be determined that he is Mr. A. Here, it is assumed that Mr. A is the only person having the key 904. If the “bell sound” is not detected before and after the “open / close lock sound” is identified, the operation of the open / close lock itself is artificial, but the key 904 is used for the operation. I have not been told. Therefore, it can be determined that the suspicious person has performed the lock. From the above, if the identification sound is “open / closed lock sound” and “bell sound” is not detected as the characteristic sound pattern, the artifact is “present”, and the person who emitted the identification sound is “suspicious person” "Is determined.

  FIG. 8 shows an example of the operation flow of the event detection apparatus 200. Differences between the operation flow of the event detection device 200 and the event detection device 100 will be described. In the event detection device 200, the human nature determination unit 240 records the identified sound types in time series, the time series generation patterns of the sound types, and the identification sounds and feature quantities registered in the human nature determination database 292. The pattern is compared and analyzed to further determine the person who has generated the identification sound in addition to the artifact (S201).

With the configuration and operation described above, the event detection apparatus 200 can further identify the person who has emitted the identification sound. That is, when a bell sound is heard at the time of opening / closing locking, it indicates that the opening / closing locking has been performed using the key 904 associated with the bell 905, and the opening / closing lock is held by Mr. A who is the owner of the key 904. Can be guessed. On the other hand, if there is no bell sound at the time of opening and closing the lock, it indicates that there is a high possibility that the opening and closing lock has been performed by some method other than the key 904. Can be guessed.
By using the identification sound as the opening / closing lock sound and the characteristic sound as the bell sound and recognizing the difference in the generation pattern of the characteristic sound by the event detection device 200, it is possible to identify the person who has performed the open / close lock. That is, it is possible to identify the person who has generated the event without forcing the user to perform a special action.

(Third embodiment)
FIG. 9 shows an example of the usage environment of the event detection apparatus 300 according to the third embodiment of the present invention. The event detection apparatus 300 can make a determination according to the occurrence pattern of the characteristic environment change in addition to the characteristic sound. Characteristic environmental change is a human action or a specific person among environmental changes that can be estimated from changes in environmental feature quantities (specifically, light quantity, temperature, air volume, moisture content, radio wave intensity, etc.) It is intended to be caused by the operation of Specifically, the “lighting of the fluorescent lamp” in the case where the occurrence of a specific light amount change at the time of lighting of the fluorescent lamp is detected is highly likely to occur based on a human operation. For this reason, “lighting of the fluorescent lamp” corresponds to a characteristic environmental change. In the present embodiment, a specific description will be given by paying particular attention to the amount of light as the environmental feature amount. The event detection device 300 may be an example of an “event detection system” in the embodiment of the present invention.

  The event detection device 300 is installed in a room in the house and can collect sound and environmental changes generated in and around the room. In this embodiment, when the door 901 is opened and closed, a sound indicating the operation is emitted. Further, when the switch of the illumination 906 is operated, a switch sound and a light amount change peculiar to the illumination are generated. In other words, the event detection apparatus 300 can detect “a change in the amount of light due to illumination” in addition to the “door open / close sound” and the “switch sound”.

  FIG. 10 shows an example of a block configuration of the event detection apparatus 300. Of the constituent elements of the event detection apparatus 300, constituent elements having the same names and having the same reference numerals as those of the event detection apparatus 100 show similar functions and operations.

  Compared to the event detection apparatus 100, the event detection apparatus 300 further includes a light amount detection element 360, a characteristic environment change identification model database 393, and a characteristic environment change identification unit 370.

  The light amount detection element 360 detects the amount of light around the event detection device 300 and converts it to a format for information processing. Specific examples of the light amount detection element 360 include an illuminance sensor, a light amount sensor, and a photodiode. Hereinafter, the above-described “format for information processing” is referred to as an environmental feature amount. The environmental feature quantity is extracted as time-series data in the same manner as the sound feature quantity vector.

  The environmental feature amount may be prepared as a combination of a plurality of data. For example, when the light quantity detection element 360 is composed of a plurality of elements and each element can detect the light quantity in a place different in real space, the light quantity obtained by each element may be treated as an independent environmental feature quantity. Good.

In the present embodiment, the amount of light is used as the environmental feature amount. On the other hand, when using other environmental feature values, it is desirable to provide an element capable of detecting the environmental feature values.
For example, if temperature is used as the environmental feature quantity, a temperature sensor, a thermal sensor, an infrared sensor, or the like is used. Further, if the radio wave intensity is used as the environmental feature amount, an antenna and a communication module that can receive the target radio wave are used. Also, a plurality of different environmental feature amounts, for example, a combination of environmental feature amounts of light quantity and temperature may be prepared.

  The characteristic environment change identification model database 393 stores data serving as a model for comparing and analyzing environmental feature amounts and specifying the presence and type of characteristic environment changes. Specific examples of model data include learned model parameters of statistical models such as feature quantities and HMMs. For both feature quantities and learned model parameters, model data is prepared for each environmental change type to be identified.

  The characteristic environment change identification unit 370 compares and analyzes the environment feature quantity and the characteristic environment change identification model database 393, and specifies the presence and type of the characteristic environment change. Also, the characteristic environment change identification unit 370 notifies the human nature determination unit 340 of the specified environment change type. Specific identification methods include likelihood determination based on statistical models such as HMM, GMM, SVM, and Bayesian estimation, and similarity determination based on distance measurement between data such as DP matching and DTW, as with the sound identification unit 130. It is done. However, the present invention is not limited to these.

  The anthropogenicity determination database 392 in the event detection apparatus 300 further holds a characteristic environment change and its generation pattern in association with the characteristic sound. Similarly, the artifact determination unit 340 performs a comparative analysis of the occurrence pattern of the characteristic environment change in addition to the identification sound and the identification sound, the characteristic sound pattern, and the occurrence pattern of the characteristic environment change registered in the artifact determination database 392. And determine human artifacts.

  FIG. 11 shows an example of data stored in the artificial determination database 392 in a table format. In the example shown in FIG. 11, when “door sound” is detected before and after the “door opening / closing sound” is identified, the door opening / closing that is the identification sound is detected. Describes that it is determined to be artificial.

  FIG. 12 shows an example of the operation flow of the event detection apparatus 300. The event detection device 300 performs the following steps S301 to S303 in addition to steps S101 to S103 for performing sound identification similar to the operation flow of the event detection device 100. Specifically, the light amount detection element 360 detects the amount of light around the event detection device 300 and converts it into an environmental feature amount (S301). The characteristic environment change identification unit 370 determines whether or not there is a characteristic environment change and the type of environment change by comparing and analyzing the obtained environment feature quantity and the characteristic environment change identification model database 393 (S302). Thereafter, the human artifact determination unit 340 records the identified sound type in the environment change type time series, the time series occurrence pattern of the sound type and the environment change type, and the identification sound registered in the humanity determination database 392. Then, the characteristic sound and the pattern of the characteristic environment change are compared and analyzed to determine the human artifact (S303). Finally, the output unit 150 performs output based on the identification sound and the presence or absence of artifacts (S105).

  With the above-described configuration and operation, the event detection apparatus 300 can determine the artifact of the event based on the environmental change in addition to the sound. Taking door opening and closing as an example, footsteps may not be detected accurately depending on the installation location. Even if the switch sound when turning on the illumination can be detected, the causal relationship between the switch sound and the door opening / closing sound cannot be understood unless the type of the switch sound is known. For this reason, the switch sound cannot be used for the determination of the artifact of the door opening / closing sound. Therefore, if a change in the amount of light that can be inferred from lighting of the lighting 906 can be detected at a timing near the door opening / closing sound and switch sound, the causal relationship between the opening / closing of the door and the lighting of the lighting is used to make an artificial opening / closing of the door Can be determined. That is, even if the sound alone cannot be used to specify the humanity or the person who emitted the sound, it is possible to specify the humanity or the person using the person-specific or person-specific environmental changes.

  The configuration and operation of the event detection apparatus in the first to third embodiments have been described above. As the characteristic sound, an electronic sound having periodicity in frequency or rhythm may be included. Specific examples of the electronic sound include a buzzer sound, an electronic melody, an operation response sound of an electronic device, and a notification sound when a call is received. Many electronic sounds are often emitted by explicit operation of electronic equipment. In addition, if the electronic sound exhibits the same operation, the characteristics of the emitted sound are almost the same every time it is generated. Therefore, it is possible to determine an artifact or a person with high accuracy.

(Fourth embodiment)
FIG. 13 shows an example of a block configuration of an event detection apparatus 400 according to the fourth embodiment of the present invention. Of the constituent elements of the event detection apparatus 400, constituent elements having the same names as those of the constituent elements of the event detection apparatus 100 show similar functions and operations. The event detection device 400 is a device capable of registering the characteristic sound acquired in the installation environment by learning. The event detection apparatus 400 may be an example of an “event detection system” in the embodiment of the present invention.

  The event detection device 400 further includes an artificial pattern learning unit 480 as compared with the event detection device 100. The artificial pattern learning unit 480 selects a feature sound candidate having (or can be determined not to have) artificiality from the generation patterns of the sound types identified by the sound identifying unit 130. Furthermore, the artificial pattern learning unit 480 associates the candidate with the determination model when the candidate is detected, and registers the candidate in the artificial determination database 192 as a characteristic sound.

  The operation of the artificial pattern learning unit 480 will be described more specifically with a bell sound as an example. It is difficult to determine whether or not the sound of a bell has an artifact by the sound of the bell itself. This is because the bell may be attached to the neck of an animal such as a cat, or may be attached to a human tool such as a key or a bag. That is, the presence or absence of artifacts in the characteristic sound changes depending on the installation environment.

  For example, if a bell sound is frequently identified before and after the occurrence of a non-artificial sound such as a cat's cry, the identification sound when the bell sound is detected before and after it may be considered to be less artificial. it can. Therefore, the artificial pattern learning unit 480 registers the bell sound in the artificial determination database 192 as having low artificiality. In addition, if bell sounds are frequently identified before and after the door opening and closing sounds and footsteps, conversely, it can be considered that the identification sound when the bell sounds are detected before and after is highly artificial. Therefore, the artificial pattern learning unit 480 registers the sound of the bell in the artificial determination database 192 as having high artificiality.

  With this configuration, the event detection apparatus 400 can select and register a sound that can be a characteristic sound from the sound types acquired in the installation environment. Examples of selection criteria for feature sound candidates include a high occurrence frequency before and after the identification sound and the characteristic sound that can be determined by human artifacts, and the occurrence of sound before and after the identification sound determined to be artificial. It is done. Therefore, the characteristic sound peculiar to the installation environment can be determined.

  It is good also as a structure further equipped with an input part and controlling the behavior of learning based on the input from a user. FIG. 14 shows an example of a block configuration of the event detection apparatus 500. Among the constituent elements of the event detecting apparatus 500, constituent elements having the same names as those of the constituent elements of the event detecting apparatus 100 or the event detecting apparatus 400 have similar functions and operations. The event detection device 500 may be an example of an “event detection system” in the embodiment of the present invention.

  The event detection device 500 further includes an input unit 590 as compared with the event detection device 400. The input unit 590 has a function of receiving input from a user and controlling the artificial pattern learning unit 480 based on the input. Further, the artificial pattern learning unit 480 determines selection of feature sound candidates and registration in the artificiality determination database 592 based on control from the input unit 590.

  By adopting such a configuration, the event detection apparatus 500 can perform a learning operation at a timing desired by the user based on an input from the user. For this reason, learning according to the convenience of the user can be performed. For example, the learning operation can be controlled by designating before and after the occurrence of the characteristic sound that the user wants to learn. In addition, when the identification sound and the characteristic sound as a result of learning include sounds that are contrary to the user's intention (for example, noise specific to the installation environment), the learning can be canceled by the user's operation. It will be possible. Therefore, learning that reflects the user's intention is possible.

  Furthermore, it is good also as a structure which designates and associates and outputs the output operation | movement at the time of detecting a characteristic sound pattern. In this case, the artificial pattern learning unit 480 further associates the designated output operation designated via the input unit 590 and registers it in the artificial nature determination database 592. The output unit 150 further performs a specified output operation registered in the anthropogenicity determination database 592 as an operation based on the determination result.

  FIG. 15 shows an example of data stored in the artificial determination database 592 in a table format. In the example illustrated in FIG. 15, it is determined that when footsteps are generated before and after the door opening / closing sound, it is determined that the operation has human artifacts, and the time and the identification sound are recorded as outputs.

  With this configuration, the event detection apparatus 500 can register the characteristic sound and the output at the time of determination in the anthropogenicity determination database 592 by user operation. Therefore, the output operation can be changed according to the importance of the identification sound or the characteristic sound. For example, it can be set to record in the storage device in the case of unimportant sound or frequent sound, to notify the user in the case of important sound, or to generate a trigger when operating other devices. .

  You may make it identify with weighting the sound generation pattern peculiar to an installation place. In this case, the artificial pattern learning unit 480 further selects and learns a sound generation pattern specific to the environment of the installation destination, and the artificiality determination unit 140 further determines by weighting the unique sound generation pattern of the installation destination. To do.

  By adopting such a configuration, the event detection apparatus 500 can determine by weighting the characteristic sound learned in the installation environment. The sound unique to the installation site is likely to reflect the habits and intentions of the user at the installation site. Therefore, it is possible to determine the artifacts with higher accuracy.

  A function may be added in which event detection devices installed in a plurality of homes are connected via a network and automatically learned based on the sound type and sound type generation pattern acquired by each event detection device. FIG. 16 shows an example of the usage environment of the event detection apparatus 600. Event detectors 600a, b, c,... Installed in a plurality of homes (hereinafter collectively referred to as event detectors 600) are connected via a network 990 and acquired by each event detector 600. Learning is automatically performed based on the sound type occurrence pattern. The event detection device 600 may be an example of an “event detection system” in the embodiment of the present invention.

  FIG. 17 shows an example of a block configuration of the event detection apparatus 600. Among the constituent elements of the event detecting apparatus 600, constituent elements having the same names as those of the constituent elements of the event detecting apparatus 100 or the event detecting apparatus 400 show similar functions and operations.

  The event detection device 600 further includes a communication unit 699 as compared with the event detection device 400. The communication unit 699 is connected to the event detection apparatus 600 other than its own apparatus via the network 101, and transmits / receives information related to the characteristic sound.

  By adopting such a configuration, the event detection apparatus 600 can compare the sound generation patterns acquired by the own apparatus and other than the own apparatus, and identify the sound generation patterns that can be generated at the installation destination. Specifically, there is a method of registering as a characteristic sound pattern when a certain sound generation pattern is hardly registered other than the device itself. The sound unique to the installation site is likely to reflect the habits and intentions of the user at the installation site. Therefore, it is possible to determine the artifacts with higher accuracy.

  When the event detection apparatus 600 determines that the feature sound registration is insufficient, a function may be added to automatically operate the learning unit to prompt the user to register the feature sound. In this case, the artificialness determination unit 140 further operates the artificial pattern learning unit 480 when it is determined that the amount of feature sound registration is insufficient for the determination by the artificiality determination unit 140 itself. Further, the output unit 150 further presents to the user that the determination of artifacts cannot be performed and that the artifact pattern learning unit 480 is operating.

  By adopting such a configuration, the event detection apparatus 600 can prompt the user to input sound so that further characteristic sounds can be learned when it is difficult to determine human artifacts. Therefore, it becomes possible to determine with higher accuracy.

  In each example shown in the fourth embodiment, the characteristic sound and the determination of the artificiality using the characteristic sound have been described. However, the characteristic environment change and the person specific determination are also described. The same applies.

  The configuration and operation of the event detection device in the first to fourth embodiments have been described above. As the characteristic sound, an electronic sound having periodicity in frequency or rhythm may be included. Specific examples of the electronic sound include a buzzer sound, an electronic melody, an operation response sound of an electronic device, and a notification sound when a call is received. Many electronic sounds are often emitted by explicit operation of electronic equipment. If the electronic sound exhibits the same operation, the characteristics of the emitted sound are substantially the same every time it is generated. Therefore, it is possible to determine an artifact or a person with high accuracy.

(Fifth embodiment)
FIG. 18 shows an example of the usage environment of the event detection apparatus 700 according to the fifth embodiment of the present invention. The event detection device 700 identifies a person who gives an instruction using electronic sound detection and activates a device / application specified in advance. The event detection device 700 may be an example of an “event detection system” in the embodiment of the present invention.

  In the present embodiment, event detection device 700, telephone 907, and personal computer 908 are installed in the house. The event detection device 700 and the personal computer 908 are connected via a network. The connection between the event detection device 700 and the personal computer 908 may be a method using USB (Universal Serial Bus), Ethernet (registered trademark), wireless LAN (Local Area Network), Bluetooth (registered trademark), or the like. The event detection device 700 is installed in a place where a ring tone from the telephone 907 can be heard. The telephone 907 is set to emit a ringtone B when receiving an incoming call from the mobile phone 909.

  FIG. 19 shows an example of data stored in the anthropogenicity determination database 792 of the event detection apparatus 700 in a table format. When the identification sound is silent, that is, when the house is absent, when the ringtone B registered as the characteristic sound is detected, the identification is made that the caller is Mr. B, and the PC 908 is used as an output via the network. Is registered to start a VPN (Virtual Private Network) connection.

  FIG. 20 shows an example of an operation sequence of the telephone set 907, the personal computer 908, the mobile phone 909, and the event detection apparatus 700. Here, it is assumed that Mr. B is a resident at the installation site, and Mr. B carries a mobile phone 909 when going out. When the user is away from home, the mobile phone 909 makes a call to the telephone 907 (S401). When the telephone set 907 receives a call, it rings the ringtone B associated with the mobile phone 909 as the caller (S402). Then, the event detection device 700 detects the ring tone B when no identification sound is input, that is, in a silent state. For this reason, the event detection apparatus 700 is registered based on the information registered in the artificial determination database 792 that the source of the information is Mr. B, and that the personal computer 908 is activated as an output to perform the VPN connection. (S123). Based on the detection result, the event detection apparatus 700 transmits a packet including the command to the personal computer 908 (S124). The personal computer 908 starts activation and VPN connection in accordance with the received command.

  By adopting such a configuration, it is possible to remotely control home devices even from outside the home. Usually, many home devices are connected to an external network via NAT (Network Address Translation) or NAPT (Network Address Port Translation) processing. For this reason, in order to control home devices directly from the outside of the network, introduction of special network devices and knowledge of the network including security are required. On the other hand, in the present embodiment, a system is provided in which a telephone ring tone, which is an electronic sound, is specified as an event detection condition, and another home device is controlled from the event detection device 700 installed in the home when the event is detected. We are using. Therefore, since the communication itself is completed at home, there is no need to be aware of security and special operations caused by the network. In addition, the telephone ringing tone serving as an event trigger is set to be generated when an incoming call is received from a specific mobile phone 909, and unauthorized operations from a third party can be prevented. Further, since the distribution and setting of the telephone ringtone is a function realized by the existing telephone 907, the introduction of new equipment can be minimized. As described above, it is possible to solve the problems of user knowledge, cost, and network security, which are problems in remote control of home devices.

  The embodiments of the present invention have been described in detail above with reference to the drawings. A program for realizing all or part of the functions of the event detection apparatus according to the embodiment of the present invention is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read into a computer system and executed. By doing so, you may process each part. The “computer system” here includes an OS and hardware such as peripheral devices.

  “Computer-readable recording medium” refers to a portable medium such as a magneto-optical disk, ROM, and nonvolatile semiconductor memory, and a storage device such as a hard disk built in a computer system. "Computer-readable recording medium" means a program that dynamically holds a program for a short time, such as a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system that serves as a server or a client in this case includes a program that holds a program for a certain period of time. The above program may realize part of the functions described above, and may further realize the functions described above in combination with a program already recorded in the computer system.

  The specific configuration is not limited to the above configuration. Various design changes can be made without departing from the scope of the present invention. For example, in the above-described embodiment, sound collection, feature extraction / identification, artificial determination, and output are all performed by the event detection device, but the present invention is not limited to this, and these processes are performed on a plurality of devices. It may be realized separately. For example, sound collection to identification can be performed with a single device, and the determination and output of human artifacts can be performed with another device, and the identification sound, characteristic sound, and similar information can be notified via the network. Good.

  As mentioned above, although this invention was demonstrated using embodiment, the technical scope of this invention is not limited to the range as described in the said embodiment. It will be apparent to those skilled in the art that various modifications or improvements can be added to the above embodiment. It is apparent from the scope of the claims that the embodiments added with such changes or improvements can be included in the technical scope of the present invention.

  (Supplementary note 1) In the event detection apparatus, the artifact determination unit further includes a generation pattern of an electronic sound having periodicity in sounds or frequencies or rhythms caused by a human action, or a learning amount of an environmental change. When it is determined that the determination unit is insufficient to make a determination, the artificial pattern learning unit is operated, and the output unit further utilizes the fact that the artificial pattern cannot be determined and the artificial pattern learning function is operating. May be presented to the person.

  This application claims the priority on the basis of Japanese application Japanese Patent Application No. 2011-025795 for which it applied on February 9, 2011, and takes in those the indications of all here.

  Since the present invention can detect an event using a sound generation pattern, it can be applied to any environment where a microphone can be installed. Moreover, although the security use in the general home was explained in mind, it is naturally applicable to other uses as well. For example, it is possible to detect an abnormal state represented by a caregiver's habit or falling by applying to a home having a care facility and a care receiver. In addition, application to hospitals, public facilities, and commercial facilities can be expected.

DESCRIPTION OF SYMBOLS 100 Event detection apparatus 110 Sound input element 120 Feature-value extraction part 130 Sound identification part 140 Artificiality judgment part 150 Output part 191 Sound identification model database 192 Artificiality judgment database 200 Event detection apparatus 240 Artificiality judgment part 292 Artificiality judgment database 300 Event detection device 340 Artificiality determination unit 360 Light amount detection element 370 Characteristic environment change identification unit 392 Humanity determination database 393 Characteristic environment change identification model database 400 Event detection device 480 Artificial pattern learning unit 500 Event detection device 590 Input unit 592 Artificial Sex determination database 600 Event detection device 699 Communication unit 700 Event detection device 792 Artificiality determination database 901 Door 902 User 903 Door lock 904 Key 905 Bell 906 Lighting 907 Telephone 908 Personal computer 909 Mobile phone 990 network

Claims (10)

An event detection system that detects an event that has occurred,
A feature quantity extraction unit that extracts a time series feature quantity vector from the signal information of the acquired sound;
A sound identification unit for identifying the type of the acquired sound by comparing and analyzing the time-series feature amount vector extracted by the feature amount extraction unit and the model data for each type of sound;
The type of acquired sound received from the sound identification unit is recorded in time series, and an artificiality determination unit that determines human artifacts based on the time series occurrence pattern of the type of acquired sound;
An event detection system comprising: an output unit that performs output in accordance with a determination result determined by the artifact determination unit. The event detection system according to claim 1, wherein the artifact determination unit further determines a person in the generation of the acquired sound based further on a sound generation pattern caused by a motion of a specific person. The event detection system according to claim 1, wherein the artifact determination unit further determines based on a generation pattern of an electronic sound having periodicity in frequency or rhythm. By comparing and analyzing the acquired environmental feature quantity and model data for each environmental change caused by a human action or a specific person's action, the presence or absence of an environmental change caused by a human action or a specific person's action, and A characteristic environment change identifying unit for identifying the type;
The event detection system according to any one of claims 1 to 3, wherein the artifact determination unit further determines based on an environment change occurrence pattern identified by the characteristic environment change identification unit. The characteristic environment change identifying unit includes the acquired environment feature amount including at least the light amount, and model data prepared for each environment change including a light amount change caused by at least a human action or a specific person action; The event detection system according to claim 4. Selection of sound generation pattern and environment change candidate caused by human action or specific person action from the identified sound and environment change occurrence pattern, and determination when the candidate is detected The event detection system according to any one of claims 1 to 5, further comprising an artificial pattern learning unit that links a model and registers the sound generation pattern and the environmental change. Further comprising an input unit that receives input from the user and controls the artificial pattern learning unit based on the input;
The event detection system according to claim 6, wherein the artificial pattern learning unit further determines selection and registration of the sound generation pattern and environmental change candidates based on control from the input unit. The artificial pattern learning unit further registers and registers the designated output operation designated through the input unit,
The event detection system according to claim 7, wherein the output unit performs the designated output operation as an operation based on a determination result. The artificial pattern learning unit selects and learns a sound generation pattern specific to an environment in which the event detection system is installed, from sounds caused by human motion or a specific human motion,
The event detection system according to any one of claims 6 to 8, wherein the artifact determination unit weights and determines the unique sound generation pattern. Generation pattern of a sound connected to an event detection device other than the event detection system via a network, a sound caused by a human action, a sound caused by a specific person action, or an electronic sound having periodicity in frequency or rhythm And a communication unit that transmits and receives information related to environmental changes,
The event detection system according to claim 6, wherein the artificial pattern learning unit selects the candidate based on information obtained through the communication unit.

Images