JP2012226595A - Gesture recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a gesture recognition device capable of limiting a period in which recognition processing on a gesture is performed without requiring any sign for the start of recognition on the gesture.SOLUTION: A detection part 15 has two operation modes, i.e., a person detection mode for detecting whether there is a person 3 in a detection range and a gesture detection mode for detecting the position of a specific part of the body of the person 3. A mode switching part 17 changes the operation mode of the detection part 15 to the gesture detection mode when the detection part 15 detects the person 3 being present during operation in the person detection mode, and stores a currently acquired image as a reference image in an area for comparison data. The detection part 15 which is operating in the gesture detection mode detects the position of the specific part of the body of the person 3 in time series by a background difference method of detecting differences between an image acquired periodically by the acquisition part 11 and the reference image stored in the area for comparison data of a storage part 12.

Description

  The present invention relates to a gesture recognition apparatus that recognizes a gesture using acquired information representing a temperature distribution within a detection range.

  Conventionally, as an example of an input interface used for control of a device to be controlled, a gesture recognition device that recognizes a gesture represented by a movement of a specific part (for example, a face) of a person has been proposed (see, for example, Patent Document 1). ).

  The gesture recognition device of Patent Literature 1 recognizes a gesture by detecting the inclination of a specific part of a person using an image taken by a camera. In the configuration of Patent Document 1, the gesture recognition process is executed when the power of the gesture recognition device is turned on.

  In addition, when the gesture recognition device is simply used as a non-contact pointing device, it only needs to be able to recognize the position of a specific part (for example, a hand), and this type of device has stable operation, low cost, and low power consumption. Such performance is required.

JP 2010-170392 A

  However, in the gesture recognition apparatus as described above, it is necessary to limit the period during which the gesture recognition process is performed in order to prevent the gesture from being recognized at an unintended timing. For this purpose, it is necessary to provide a gesture recognition start signal as in the case of powering on the gesture recognition device in Patent Document 1, and therefore it is necessary to provide an interface different from the gesture input in the gesture recognition device. There is a problem that cost reduction is difficult.

  The present invention has been made in view of the above-described reasons, and an object thereof is to provide a gesture recognition device that does not require a gesture recognition start signal and can limit a period during which gesture recognition processing is performed.

  The gesture recognition device according to the present invention includes an acquisition unit that acquires acquisition information representing the heat flux distribution from a sensor device that detects the distribution of heat flux received from the detection range, and the detection when no person is present in the detection range. Two operation modes: a storage unit that stores background information representing the distribution of heat flux within the range, a human detection mode that detects the presence or absence of a person within the detection range, and a gesture detection mode that detects the position of a specific part of the person A recognition unit that recognizes a gesture based on the position of the specific part detected in the gesture detection mode, and a mode switching unit that switches an operation mode of the detection unit, and the mode switching unit includes: When the detection unit detects the presence of a person during operation in the person detection mode, the acquired information at the time when the person is detected is used as reference information, and the operation mode of the detection unit is detected. The gesture detection mode is switched to the gesture detection mode, and the detection unit detects the presence / absence of a person in the detection range based on a difference between the acquired information and the background information in the person detection mode, and detects the gesture In the mode, the position of the specific part is detected based on a difference between the acquired information and the reference information.

  In this gesture recognition apparatus, when the mode switching unit detects that a person has left the detection range based on a change in the acquired information while the detection unit is operating in the gesture detection mode, the operation of the detection unit It is desirable to switch the mode to the human detection mode.

  The gesture recognition apparatus further includes an update unit that updates the background information at a predetermined update time interval while the detection unit is operating in the human detection mode, and the update unit is configured to start and end the gesture detection mode. Is longer than the update time, it is more desirable to update the background information with the acquired information at the end of the gesture detection mode.

  The present invention does not require a signal for starting gesture recognition, and has an advantage that the period during which gesture recognition processing is performed can be limited.

It is the schematic which shows the structure of the gesture recognition apparatus which concerns on Embodiment 1. FIG. It is explanatory drawing showing the whole process of the gesture recognition apparatus which concerns on Embodiment 1. FIG. It is explanatory drawing showing the human approach detection process of the gesture recognition apparatus which concerns on Embodiment 1. FIG. It is explanatory drawing showing the specific site | part detection process of the gesture recognition apparatus which concerns on Embodiment 1. FIG. It is explanatory drawing showing the person leaving detection process of the gesture recognition apparatus which concerns on Embodiment 1. FIG. It is explanatory drawing which shows operation | movement of the gesture recognition apparatus which concerns on Embodiment 1. FIG. It is explanatory drawing showing the whole process of the gesture recognition apparatus which concerns on Embodiment 2. FIG.

(Embodiment 1)
As shown in FIG. 1, the gesture recognition device 1 according to the present embodiment includes an acquisition unit 11 that acquires information from the sensor device 2, a storage unit 12 that stores information, and processes such as gesture recognition by performing various calculations. And an output unit 14 that outputs a recognition result.

  The sensor device 2 is configured by two-dimensionally arranging sensor elements (not shown) having sensitivity to infrared rays, such as a thermopile element, in a matrix, and information indicating the distribution of heat flux received from the detection range by each sensor element. Output continuously as a two-dimensional thermal image. Each sensor element corresponds to each pixel of the thermal image, and hereinafter, the Celsius temperature of the area detected by each sensor element will be described as the pixel value of each corresponding pixel in the thermal image. In the example of FIG. 1, the gesture recognition device 1 is separate from the sensor device 2. However, the present invention is not limited to this configuration, and the gesture recognition device 1 may be integrally provided with the sensor device 2.

  The gesture recognition device 1 according to the present embodiment includes a computer. By causing the computer to execute a predetermined program, the functions of the acquisition unit 11 and the arithmetic processing unit 13 are realized by the computer. The output unit 14 includes a device having a function of presenting the recognition result of the arithmetic processing unit 13 to the user, such as a monitor device or a printer device.

  The acquisition unit 11 periodically acquires an acquired image, which is a thermal image representing the heat flux distribution in the current detection range, from the sensor device 2 as acquisition information. The gesture recognition device 1 stores a thermal image representing the heat flux distribution in the detection range at the time when the person 3 does not exist in the detection range as a background image (background information) in advance. In the present embodiment, the storage unit 12 has an acquisition data area for storing an acquired image, a comparison data area for storing an image to be compared with the acquired image, and a return data area to be described later. In the comparison data area, a background image is stored as an initial value.

  The arithmetic processing unit 13 includes a detection unit 15 that detects the presence or absence of the person 3 within the detection range and the position of a specific part of the body of the person 3, a recognition unit 16 that recognizes a gesture based on the position of the specific part, and a detection unit And a mode switching unit 17 for switching 15 operation modes.

  The detection unit 15 has two operation modes: a person detection mode for detecting the presence or absence of the person 3 within the detection range, and a gesture detection mode for detecting the position of a specific part of the body of the person 3.

  When the detection unit 15 detects the presence of the person 3 while operating in the human detection mode, the mode switching unit 17 switches the operation mode of the detection unit 15 from the human detection mode to the gesture detection mode. That is, when detecting the presence of the person 3 within the detection range in the person detection mode, the detection unit 15 ends the person detection mode and starts the gesture detection mode. When switching to the gesture detection mode, the mode switching unit 17 copies the data (background image) from the comparison data area of the storage unit 12 to the return data area, and at that time (when the person 3 is detected). The acquired image is stored in the comparison data area as a reference image (reference information).

  On the other hand, when the mode switching unit 17 detects that the person 3 has left the detection range while the detection unit 15 is operating in the gesture detection mode, the mode switching unit 17 switches the operation mode of the detection unit 15 from the gesture detection mode to the human detection mode. That is, when it becomes impossible to detect the person 3 within the detection range in the gesture detection mode, the detection unit 15 ends the gesture detection mode and starts the person detection mode. When switching to the human detection mode, the mode switching unit 17 copies data (background image) from the return data area of the storage unit 12 to the comparison data area.

  In short, when switching from the human detection mode to the gesture detection mode, the mode switching unit 17 temporarily saves the background image stored in the comparison data area in the return data area and stores the reference image in the comparison data area. Set. When switching from the gesture detection mode to the human detection mode, the mode switching unit 17 deletes the reference image stored in the comparison data area and temporarily saves the reference image in the return data area. Is returned to the comparison data area. As a result, the background image is set in the comparison data area of the storage unit 12 while the detection unit 15 is operating in the human detection mode, and the reference image is set during operation in the gesture detection mode.

  While operating in the human detection mode, the detection unit 15 detects a difference between the acquired image periodically acquired by the acquisition unit 11 and the background image stored in the comparison data area of the storage unit 12. The presence or absence of the person 3 within the detection range is detected by the background subtraction method. That is, in the human detection mode, the detection unit 15 detects the approach of the person 3 to the sensor device 2 by detecting that the person 3 has entered the detection range of the sensor device 2 (hereinafter, “ Human approach detection process).

  On the other hand, the detection unit 15 calculates the difference between the acquired image periodically acquired by the acquisition unit 11 and the reference image stored in the comparison data area of the storage unit 12 during operation in the gesture detection mode. The position of the specific part of the body of the person 3 is detected in time series by the background difference method to be detected. That is, in the gesture detection mode, the detection unit 15 detects the position of a specific part where the heat flux value is further increased against the background of the heat flux distribution caused by the person 3 existing in the detection range. Processing (hereinafter referred to as “specific part detection processing”) is performed.

  In the present embodiment, as an example, the part from the wrist of the person 3's hand is the specific part. If the person 3 puts out a hand as a specific part over the sensor device 2, the heat flux value received by the sensor device 2 when the specific part approaches the sensor device 2 is changed to other parts of the body of the person 3. It becomes larger in a specific part compared with. Therefore, when the detection unit 15 is operating in the gesture detection mode, if the person 3 puts out a hand as a specific part, the heat flux value in the acquired image is larger in the specific part than in other parts of the body. Become. Accordingly, when the person 3 approaches the sensor device 2 and the operation mode of the detection unit 15 is switched from the person detection mode to the gesture detection mode, the person 3 projects the hand toward the sensor device 2, thereby The position of the hand as can be detected by the detection unit 15. The detection unit 15 outputs the barycentric coordinates of the specific part (here, the part ahead of the wrist of the hand) detected as the heat source as the position of the specific part.

  The recognition unit 16 recognizes a gesture based on the position of the specific part (here, the part from the wrist of the hand) detected by the detection unit 15 in the gesture detection mode. The storage unit 12 stores in advance a correspondence table that represents a correspondence relationship between a plurality of types of gestures and the position of a specific part or a change in the position of the specific part over time (that is, “movement”). The recognizing unit 16 extracts the position of the specific part or its movement from the detection result of the detecting unit 15, and recognizes the gesture by comparing the extracted result with the correspondence table. In addition, about the method of recognizing a gesture from the position of a specific site | part, since a well-known method is applicable, detailed description is abbreviate | omitted.

  Further, the detection unit 15 calculates a difference between the acquired image periodically acquired by the acquisition unit 11 and the reference image stored in the comparison data area of the storage unit 12 during operation in the gesture detection mode. The presence or absence of the person 3 within the detection range is also detected by the background difference method to be detected. That is, in the gesture detection mode, the detection unit 15 detects that the person 3 has left the detection range by detecting that the person 3 is no longer in the detection range of the sensor device 2 (hereinafter “human”). "Exit detection processing").

  In the following, regarding the specific operation of the gesture recognition device 1, FIG. 2 showing the entire processing of the gesture recognition device 1, FIG. 3 showing the human approach detection processing, FIG. 4 showing the specific part detection processing, and human exit detection processing are shown. This will be described with reference to each flowchart of FIG. In the following, the pixel value of each pixel in the image is specified by the coordinate position (x, y) of each pixel in the image and time t.

  The gesture recognition device 1 first acquires a thermal image output from the sensor device 2 in the absence of the person 3 by the acquisition unit 11, and sets each pixel value of the thermal image as comparison data Ta (x, y, The initial value of t) is stored as an initial value in the comparison data area of the storage unit 12 (S1 in FIG. 2). The comparison data Ta (x, y, t) is a pixel value of each pixel constituting the comparison image Ta to be compared with the acquired image T. The comparison image Ta corresponds to a background image if the operation mode of the detection unit 15 is the human detection mode, and corresponds to a reference image if the operation mode of the detection unit 15 is the gesture detection mode. Further, in the initial setting (S1), the mode switching unit 17 sets the operation mode of the detection unit 15 to the human detection mode.

  The gesture recognition device 1 acquires the current temperature distribution data T (x, y, t) from the sensor device 2 by the acquisition unit 11 and stores the temperature distribution data T (x, y, t) in the storage unit 12. (S2 in FIG. 2). The temperature distribution data T (x, y, t) is a pixel value of each pixel constituting the acquired image T.

  The detection unit 15 calculates the difference between the comparison data Ta (x, y, t) and the temperature distribution data T (x, y, t) stored in the storage unit 12 as difference data Tb (x, y, t). (S3 in FIG. 2). That is, the difference data Tb (x, y, t) is expressed by Equation 1. The difference data Tb (x, y, t) is a pixel value of each pixel constituting the difference image Tb. The detection unit 15 performs the calculation expressed by Equation 1 for all pixels of the acquired image T for each pixel.

  Then, if the operation mode is not the gesture detection mode (S4: No in FIG. 2), the detection unit 15 performs a human approach detection process (S5). Details of the person approach detection process (S5) will be described with reference to FIG.

  That is, in the human approach detection process, the detection unit 15 includes difference data Tb (x, y, t) constituting a difference image Tb that is a difference between the acquired image T and the comparison image (here, the background image) Ta. Is binarized by a predetermined first threshold (S51 in FIG. 3). At this time, the detection unit 15 sets a pixel having a difference value larger than the first threshold as a “detection pixel” in which the heat source is detected, and detects a pixel in which the difference value is equal to or less than the first threshold as “no detection” The detection image Tc is generated by binarizing so as to obtain “pixel”. A pixel value Tc (x, y, t) of each pixel constituting the detection image Tc is expressed by Equation 2 using the first threshold C1.

  The detection unit 15 obtains, from the detection image Tc, an area of a detection region including a set of detection pixels, that is, a set of pixels with Tc (x, y, t) = 1, as a heat source detection area s1 (S52 in FIG. 3). ). Note that the pixel (x, y) where Tc (x, y, t) = 0 is a non-detected pixel. If the heat source detection area s1 is equal to or smaller than the predetermined second threshold value (S53: No in FIG. 3), the detection unit 15 determines that there is no person 3 within the detection range (S54). If the heat source detection area s1 is larger than the predetermined second threshold (S53: Yes in FIG. 3), the detection unit 15 determines that the person 3 is within the detection range, that is, the person 3 has approached the sensor device 2 ( S55).

  If the person 3 is not detected in the above-described person approach detection process (S6: No in FIG. 2), the process returns to S2. On the other hand, if the person 3 is detected (S6: Yes in FIG. 2), the mode switching unit 17 updates the return image Td in the return data area with the comparison image Ta in the comparison data area (S7). The comparison image Ta in the comparison data area is updated with the acquired image T at that time (S8). In other words, the mode switching unit 17 copies the background image from the comparison data area of the storage unit 12 to the return data area (S7), and stores the acquired image T at that time as a reference image in the comparison data area. (S8). Then, the mode switching unit 17 switches the operation mode of the detection unit 15 to the gesture detection mode (S9 in FIG. 2), and returns to the process of S2.

  If the operation mode is the gesture detection mode (S4: Yes in FIG. 2), the detection unit 15 executes the specific part detection process (S10). Details of the specific part detection process (S10) will be described with reference to FIG.

  That is, in the specific part detection process, the detection unit 15 uses the difference data Tb (x, y, t) constituting the difference image Tb that is the difference between the acquired image T and the comparison image (here, the reference image) Ta. Is binarized by a predetermined first threshold (S101 in FIG. 4). At this time, the detection unit 15 sets a pixel having a difference value larger than the first threshold as a “detection pixel” in which the heat source is detected, and detects a pixel in which the difference value is equal to or less than the first threshold as “no detection” The detection image Tc is generated by binarizing so as to obtain “pixel”.

  The detection unit 15 performs a labeling process on the detection image Tc, divides a detection region including a set of detection pixels, that is, a set of pixels where Tc (x, y, t) = 1, into areas each having a predetermined area. Is deleted (S102 in FIG. 4). At this time, if a region having an area equal to or larger than the predetermined value does not remain, the detection unit 15 determines that there is no heat source region (S103: No in FIG. 4), and detects the specific part as no specific part position information (S104). The process ends. On the other hand, when a region having an area equal to or larger than the predetermined value remains, the detection unit 15 determines that there is a heat source region (S103: Yes in FIG. 4), and calculates the position of the remaining region (heat source region). At this time, the detection unit 15 takes the weighted average of the coordinate values of all the pixels constituting the heat source region, thereby obtaining the gravity center position of the heat source region (Gx if one-dimensional, [Gx, Gy] if two-dimensional. ) Is calculated (S105 in FIG. 4). The detection unit 15 may easily obtain the center-of-gravity position with all the weights set to 1, and may use the temperature or the difference temperature value of each pixel that is appropriately offset as the weight. The detection unit 15 outputs the position of the center of gravity obtained in this way as the position of the specific part of the person 3 (S106 in FIG. 4).

  When the specific part detection process ends, the detection unit 15 outputs the detected position of the specific part to the recognition unit 16 (S11 in FIG. 2), and executes a person exit detection process (S12). Details of the person leaving detection process (S12) will be described with reference to FIG.

  In other words, in the person leaving detection process, the detection unit 15 includes difference data Tb (x, y, t) constituting a difference image Tb that is a difference between the acquired image T and the comparison image (here, the background image) Ta. Is binarized by a predetermined third threshold (S121 in FIG. 5). At this time, the detection unit 15 sets a pixel having a difference value smaller than the third threshold as a “non-detection pixel” in which the heat source has not been detected, and detects a pixel having a difference value equal to or greater than the third threshold as the “detection”. It binarizes so as to be “pixel”, and generates an undetected image Te. The pixel value Te (x, y, t) of each pixel constituting the undetected image Te is expressed by Equation 3 using the third threshold C3. Here, since the person leaving detection process is a process for detecting that the person 3 is gone, the third threshold C3 is a negative value.

  The detection unit 15 obtains, from the non-detected image Te, an area of a non-detection area formed of a set of non-detection pixels, that is, a set of pixels with Te (x, y, t) = 1, as a non-detection area s0 (FIG. 5 S122). Note that a pixel (x, y) where Te (x, y, t) = 0 is a detection pixel. If the non-detection area s0 is equal to or smaller than the predetermined fourth threshold (S123 in FIG. 5: No), the detection unit 15 determines that there is a person 3 within the detection range, that is, no person 3 has left the detection range. (S124). If the non-detection area s0 is larger than the predetermined fourth threshold (S123 in FIG. 5: Yes), the detection unit 15 determines that there is no person 3 within the detection range, that is, that the person 3 has left the detection range ( S125).

  If the exit of the person 3 is not detected in the person exit detection process described above (S13: No in FIG. 2), the process returns to the process of S2. On the other hand, when the exit of the person 3 is detected (S13: Yes in FIG. 2), the mode switching unit 17 updates the comparison image Ta in the comparison data area with the return image Td in the return data area ( S14). In other words, the mode switching unit 17 copies the background image from the return data area of the storage unit 12 to the comparison data area. Then, the mode switching unit 17 switches the operation mode of the detection unit 15 to the human detection mode (S15 in FIG. 2), and returns to the process of S2.

  Next, a method for recognizing a gesture using the gesture recognition device 1 having the above-described configuration will be described with reference to FIG. FIG. 6 exemplifies the process of the gesture recognition apparatus 1 in each of the scenes (a) to (f) arranged in time series, taking a 6 × 6 pixel thermal image as an example.

  First, in the first scene (a), since the operation mode of the detection unit 15 is the human detection mode and the person 3 is not in the detection range of the sensor device 2, the comparative image Ta is the background image, and the acquired image In T, the distribution of heat flux due to the person 3 does not appear.

  In the scene (b), when the person 3 approaches the sensor device 2, the heat flux distribution due to the person 3 appears in the acquired image T. Therefore, the gesture recognition device 1 uses the detection unit 15 to detect the person 3. Detect presence. At this time, the mode switching unit 17 updates the return image Td with the comparison image Ta, updates the comparison image Ta with the acquired image T at that time, and switches the operation mode of the detection unit 15 to the gesture detection mode. .

  In the next scene (c), since the person 3 raises a specific part (here, the part ahead of the wrist of the hand), a heat source corresponding to the specific part appears in the acquired image T. The position of the specific part is detected by the detection unit 15. Furthermore, in the scene (d), the position of the heat source in the acquired image T changes (descends) when the person 3 lowers a specific part (here, the part ahead of the wrist of the hand). Then, the detection unit 15 detects a change in the position of the specific part. The recognition unit 16 recognizes a gesture based on the position of the specific part detected by the detection unit 15 in the scenes (c) and (d).

  In the next scene (e), when the person 3 moves away from the sensor device 2, the distribution of the heat flux in the acquired image T changes, so that the gesture recognition device 1 detects the exit of the person 3 by the detection unit 15. To do. At this time, the mode switching unit 17 updates the comparison image Ta with the return image Td, and switches the operation mode of the detection unit 15 to the human detection mode.

  In the last scene (f), since the operation mode of the detection unit 15 is the human detection mode, and the person 3 is not in the detection range of the sensor device 2, the comparison image Ta is the background image, and the acquired image T The distribution of heat flux due to person 3 does not appear.

  According to the gesture recognition device 1 of the present embodiment described above, when the detection unit 15 detects the presence of the person 3 while operating in the human detection mode, the operation mode of the detection unit 15 is the gesture detection mode for performing gesture recognition processing. Will automatically switch to. That is, in the gesture recognition device 1, the gesture recognition process is performed only in the gesture detection mode, and the switching from the person detection mode to the gesture detection mode is triggered by the approach of the person 3 to the sensor device 2 as a trigger. Is done by. Therefore, the gesture recognition device 1 does not require a signal for starting gesture recognition, and can limit the period during which the gesture recognition process is performed.

  In short, the gesture recognition device 1 does not recognize a gesture in the human detection mode, and accepts a gesture input only when the person 3 approaches the sensor device 2 up to a certain distance, so that the gesture is recognized at an unintended timing. Can be prevented. Note that a method for preventing the recognition of an unintended gesture by complicating the recognized gesture itself with respect to the configuration of the present embodiment is also conceivable, but this method requires complicated gesture recognition processing. Therefore, it is difficult to reduce the cost of the apparatus.

  In addition, in the gesture recognition device 1 of the present embodiment, in the gesture detection mode, the acquired image in a state in which the person 3 approaches the sensor device 2 is used as a comparison image (reference image). The site can be reliably detected as a heat source. Therefore, according to this gesture recognition device 1, stable gesture input can be realized.

  Furthermore, when the person 3 leaves the detection range, the mode switching unit 17 switches the operation mode of the detection unit 15 to the person detection mode, and thus interrupts the gesture input until the next person 3 is detected. Thus, it can be detected that the person 3 is approaching anew. That is, the gesture recognition device 1 does not require a signal for ending the gesture recognition, and can limit the period during which the gesture recognition process is performed.

(Embodiment 2)
The gesture recognition device 1 of the present embodiment includes an update processing unit (not shown) that periodically updates a background image (background information) in the storage unit 12 during a period in which the detection unit 15 is in the human detection mode. 13 is different from the gesture recognition device 1 of the first embodiment. Hereinafter, the same configurations as those of the first embodiment are denoted by common reference numerals, and description thereof will be omitted as appropriate.

  That is, if the background image set in advance is used as it is, the gesture recognition device 1 causes the temperature within the detection range that does not depend on the person 3 (hereinafter referred to as “environment temperature”) due to a change in time zone or air conditioning. There is a possibility of false detection when it changes. Therefore, the gesture recognition device 1 of the present embodiment includes an update unit that repeatedly updates the background image stored in the storage unit 12 at predetermined update time intervals while the detection unit 15 is in the human detection mode. .

  The update unit measures time by a timer (not shown) during the operation of the gesture recognition device 1, and when the update time elapses when the gesture recognition device 1 is activated or the previous background image is updated, at that time. If the detection unit 15 is in the human detection mode, the background image is periodically updated. Note that the update unit may be configured to measure the update time by counting the number of frames of the acquired image T that the acquisition unit 11 periodically acquires from the sensor device 2.

  Periodic updating of the background image by the updating unit is performed as follows using the current acquired image T and the update coefficient α. That is, the update unit weighted average (weighted average) of the comparison image Ta (t−1) before regular update stored in the comparison data area of the storage unit 12 and the current acquired image T (t). The background image is regularly updated by calculating the value. That is, the comparative image (t) after the periodic update is expressed by Equation 4 using the update coefficient α which is a weighted average parameter. The update coefficient α is a constant that appears when 0 <α ≦ 1.

  By the way, the update unit measures the time by a timer during the operation of the gesture recognition device 1, and the detection unit 15 performs the gesture when the update time has elapsed since the start of the gesture recognition device 1 or the previous background image update. If in the detection mode, the periodic update of the background image is suspended. In this case, after the detection unit 15 detects the exit of the person 3 and the mode switching unit 17 replaces the comparison image Ta in the comparison data area with the return image Td in the return data area, the update unit compares the comparison image Ta. The image (here, the background image) Ta is updated. That is, when the time from the start to the end of the gesture detection mode is longer than the update time, the update unit updates the background image with the acquired image T at the end of the gesture detection mode.

  Below, operation | movement of the gesture recognition apparatus 1 of this embodiment is demonstrated with reference to the flowchart of FIG. The flowchart of FIG. 7 is different from the flowchart of FIG. 2 in that the initialization of the timer value Ut at the initial setting (S1) (to zero) and the processing of S21 to 27 are added. The processing is the same as in FIG.

  The timer increments the timer value Ut between the calculation process of the difference image Tb (S3 in FIG. 7) and the determination process whether the operation mode of the detection unit 15 is the gesture detection mode (S4) (S21). .

  In the person detection mode (S4: No), when the person 3 is not detected in the person approach detection process (S5) (S6: No), if the timer value Ut is equal to or longer than the update time (S22: Yes), the update is performed. The unit periodically updates the comparative image Ta (S23), and resets the timer value Ut (S24). If the timer value Ut is not equal to or longer than the update time (S22: No), the update unit returns to the process S2 without performing periodic update.

  In the gesture detection mode (S4: Yes), when the exit of the person 3 is detected in the human body exit process (S12) (S13: Yes), the mode switching unit 17 replaces the comparison image Ta with the return image Td. After (S14), the update unit compares the timer value Ut with the update time (S25). If the timer value Ut is equal to or longer than the update time (S25: Yes), the update unit periodically updates the comparison image Ta (S26), resets the timer value Ut (S27), and the operation mode of the detection unit 15 is performed. The process proceeds to the switching process (S15). If the timer value Ut is not equal to or longer than the update time (S25: No), the update unit proceeds to the process S15 without performing regular update.

  According to the gesture recognition device 1 of the present embodiment described above, in the human detection mode, the comparison image (here, the background image) Ta stored in the storage unit 12 is updated at a predetermined update time interval. A false detection can be avoided when the environmental temperature changes. In addition, when the gesture detection mode continues for a long time exceeding the update time, the comparison image Ta is updated at the end of the gesture detection mode, so that the gesture recognition device 1 is operating in the gesture detection mode. It is possible to respond quickly to changes in the environmental temperature that occur.

  Other configurations and functions are the same as those of the first embodiment.

DESCRIPTION OF SYMBOLS 1 Gesture recognition apparatus 3 Person 11 Acquisition part 12 Storage part 15 Detection part 16 Recognition part 17 Mode switching part T Acquisition image (acquisition information)
Ta comparison image (reference image, background image)

Claims (3)

An acquisition unit that acquires acquisition information representing the distribution of the heat flux from a sensor device that detects a distribution of heat flux received from the detection range, and a distribution of heat flux within the detection range when no person is present in the detection range. A detection unit having two operation modes: a storage unit for storing background information to be represented; a human detection mode for detecting the presence or absence of a person within the detection range; and a gesture detection mode for detecting the position of a specific part of the person; and the gesture A recognition unit that recognizes a gesture based on the position of the specific part detected in the detection mode, and a mode switching unit that switches an operation mode of the detection unit,
When the detection unit detects the presence of a person during operation in the person detection mode, the mode switching unit uses the acquired information at the time when the person is detected as reference information, and sets the operation mode of the detection unit as the gesture. Switch to detection mode,
In the person detection mode, the detection unit detects the presence or absence of a person within the detection range based on a difference between the acquisition information and the background information. In the gesture detection mode, the detection unit and the acquisition information A gesture recognition device that detects a position of the specific part based on a difference from reference information.   When the mode switching unit detects that a person has left the detection range based on a change in the acquired information while the detection unit is operating in the gesture detection mode, the mode switching unit sets the operation mode of the detection unit to the human detection mode. The gesture recognition device according to claim 1, wherein the gesture recognition device is switched to. An update unit that updates the background information at a predetermined update time interval while the detection unit is operating in the human detection mode,
The update unit updates the background information with the acquired information at the end of the gesture detection mode when the time from the start to the end of the gesture detection mode is longer than the update time. The gesture recognition device according to claim 1 or 2.

Images