JP2007047671A - Pattern recognition device and pattern recognition method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent an excessive setting value that can affect subsequent operations from being set, when the user sets the operating parameters of a pattern recognition utilizing pattern recognition. <P>SOLUTION: When a user's input is decided as an operation parameter change command of a pattern recognition by the pattern recognition, pattern recognition is conducted again for the user's input which has been held, using the operation parameter changed. If the results of the two pattern recognitions are different, it is decided that the operation parameter of the pattern recognition set by the user is an excessive value, and warning information is presented to the user. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an apparatus and method for inputting information such as voice and handwritten characters and recognizing the pattern, and a medium recording the method.

In recent years, the technology for pattern recognition has been remarkably developed, and it has been installed in various products.
General pattern recognition is a process in which a feature parameter input by a user is compared with a standard pattern (recognition target) collected in advance, and a standard pattern with the highest likelihood score is output as a recognition result. . The information input using pattern recognition is characterized by being tolerant to a certain degree of “input fluctuation”, unlike a uniform input such as text input via a keyboard.

  For example, there is a technique called voice recognition that recognizes a voice pattern. With this recognition technology, it is possible to perform recognition by absorbing the utterance of a speaker different from the speaker who collected the standard pattern, a certain amount of stagnation, and a difference in accent. As described above, the pattern recognition technology capable of absorbing “input fluctuation” is an important basic element for providing an input device that can be easily operated even by a user who is unfamiliar with devices such as children and the elderly.

  Currently used pattern recognition technologies are: speech recognition technology for recognizing speech, handwritten character recognition for recognizing dynamic / static handwritten characters, and gesture recognition for recognizing movements of the hand and mouse trajectories. And biometrics recognition for recognizing a user's physical features such as face and iris. Some of these have already been commercialized.

Moreover, as a prior art example, patent document 1 and patent document 2 can be mention | raise | lifted, for example.
JP 2005-117541 A JP 2003-67167 A

  Here, speech recognition is considered as an example of pattern recognition processing.

  The user can input a command for operating the device, such as “right” or “start copying”, while uttering a setting command related to the operation of the pattern recognition process, such as “set branching parameter to 20”. Of course it is also possible.

  The branch picking process given here as an example is a technique for speeding up the process used in pattern recognition. How much this branch hunting process is performed can be controlled by a numerical value called a branch hunting parameter. In general, when the branch picking process is strongly applied, the recognition speed is improved, but the recognition accuracy tends to be lowered. For this reason, if an excessive value is set, subsequent user input may not be successful at all.

  For setting commands that have a fatal effect on the operation of such a system, it is common to confirm the user using the following dialogue.

User: “Set branch picking parameter to 20”
System: “Is it possible to change the branch picking parameter to 20?
User: “Yes”
Taking confirmation in such a dialogue is effective in the sense of preventing an excessive setting. However, depending on the voice quality of the user and the environment in which the device is placed, the value at which the trouble occurs varies. For example, when the user's voice quality matches the standard pattern and the environmental noise is relatively low, even if a strong branch hunting process is performed, the subsequent recognition hardly occurs. On the other hand, when the user's voice quality is far from the standard pattern or in an environment where there are many additive / multiplicative noises, there is a possibility that the subsequent recognition operation will not succeed at all even with a modest branch hunting parameter.

  In this way, it is impossible to determine up to which value of the pattern recognition parameter is safe at the stage of designing the apparatus. In view of this, there is no prior art that allows a user to set a pattern recognition parameter.

  A pattern recognition apparatus according to the present invention includes a user input information capturing unit that captures user input information, a user input information storage unit that stores the user input information, and a pattern recognition parameter that stores a pattern recognition parameter that determines a pattern recognition operation. Storage means, first pattern recognition means for recognizing the user input information using the pattern recognition parameter, and converting it to a first command, output means for outputting a first command, and the first command In the case of a command for changing the pattern recognition parameter, the user recognition information stored in the user input information storage means is recognized again and converted into a second command using the pattern recognition parameter according to the command. From the second pattern recognition means, the comparison means for comparing the first command and the second command That.

  In the present invention, when the user input is a pattern recognition parameter change command, the user input is recognized again after the parameter change, and a warning is given to the user when the recognition results before and after the parameter change differ. It becomes possible to output. This makes it possible to determine whether the set pattern recognition parameters are excessive.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a functional configuration diagram of the pattern recognition apparatus of the present invention. FIG. 2 is an operation flow of the pattern recognition apparatus of this patent. Here, it demonstrates using these figures. For convenience of explanation, voice recognition is described as an example of pattern recognition processing.

  The pattern recognition device (101) of the present invention is connected to an input device (102), and takes in input information from the user. The input device may be a microphone, a mouse, a camera, or the like. The user input information may be voice information uttered by the user, moving picture information in which the user's movement is captured, or the like. The input user input information is captured from the user input capturing unit (106) into the apparatus (S201).

  The captured user input information is sent to the user input storage unit (103) and the first pattern recognition unit (107). The user input storage unit stores the user input information as user input information (104) in a storage area in the apparatus (S202). In the case of speech recognition, user utterance waveform data or feature parameters after acoustic analysis are stored.

  On the other hand, the first pattern recognition unit recognizes the transmitted user input information using the pattern recognition parameter (105) stored in the apparatus (S203). The pattern recognition at this time uses an existing technique, and thus the description thereof is omitted. In general, a probability density distribution or a neural network is often used. The pattern recognition parameters are various parameters used when performing pattern recognition, and various constants used in the pattern recognition processing algorithm correspond to this. For example, a branch hunting threshold value used in the pattern recognition process is one of the pattern recognition parameters. Branch picking is a process that cancels the calculation of a candidate that seems unlikely, and contributes to a reduction in the amount of calculation for pattern recognition.

  When the recognition process is completed in the first pattern recognition unit, it is determined whether the recognition result is a pattern recognition parameter change command (S204). If this is not related to the pattern recognition parameter, the recognition result is sent to the command output unit (109). For example, when the user input (user utterance) is recognized as “copy start”, “move right”, etc., this is not related to the parameter for pattern recognition. Therefore, the recognition result is output from the command output unit to the outside of the apparatus (S205).

  If the recognition result of the first pattern recognition unit is a pattern recognition parameter change command, the result is sent to the second pattern recognition unit (108). For example, when the recognition result is “Change the branch hunting threshold to 40”, the recognition result relates to the pattern recognition parameter. Therefore, this recognition result is sent to the second pattern recognition unit.

  The second pattern recognition unit receives the pattern recognition parameter change command, which is the recognition result of the first pattern recognition unit, and temporarily changes the pattern recognition parameter. Using the above example, temporarily change the branch picking threshold to 40. Then, as the pattern recognition parameters other than the branch hunting threshold, those stored in the device are used as they are, and the user input information (104) stored earlier is again recognized as a pattern (S206).

  When the first pattern recognition and the second pattern recognition are completed, the respective recognition results are sent to the command comparison unit (110). Here, the recognition results from the two recognition units are compared, and it is determined whether each result is equal (S207). If these recognition results are different, it can be determined that the pattern recognition parameter change command input by the user is excessive for the user input. If such a pattern recognition parameter is set, at least the current recognition has not been successful, and there is a possibility that the next processing will not be successful. However, it is possible to determine whether or not the user-recognized pattern recognition parameter is an excessive value by trying once before changing the pattern recognition parameter.

FIG. 3 is a functional configuration diagram of the pattern recognition apparatus according to the second embodiment of the present invention. Here, the same parts as those in FIG. 1 are assigned the same numbers, and the description of those parts is omitted.
As described above, when the command comparison unit (110) determines that the recognition results of the first pattern recognition unit (107) and the second pattern recognition unit (108) are different, the pattern input by the user is determined. It is determined that the recognition parameter change command is excessive for the input. In that case, the warning output unit (302) of the apparatus is notified of this, and warning information is output via the output device (301) outside the apparatus such as a display and a speaker. For example, a voice message such as “If the branch picking threshold is changed to 40, the recognition performance from the next time may be affected” is sent from a speaker.

  If the command comparison unit determines that the recognition results of the first pattern recognition unit (107) and the second pattern recognition unit (108) are equal, a message to that effect is sent to the pattern recognition parameter change unit (303). It is done. Here, the pattern recognition parameters are changed according to the recognition result of the first pattern recognition unit. In the previous example, the branch hunting threshold is set to a value of 40.

  As a result, if the pattern recognition parameter change command input by the user is an excessive value, a warning can be issued to the user.

  4A and 4B show an example in which the pattern recognition apparatus of the present invention is mounted on a portable information terminal.

  The pattern recognition performed by this apparatus is online handwritten character recognition and receives online handwritten information as user input information. The portable information terminal (401) is equipped with a touch panel (403), and the user inputs handwritten character information (404) using a stylus (402) or the like.

  The user can operate applications such as a scheduler and a mailer by using handwritten character input or GUI. Further, the operation for recognizing handwritten characters can be changed by a recognition parameter changing command. For example, as shown in FIG. 4A, the apparatus can be notified that the speed of handwritten character recognition is to be increased.

  In the apparatus, this handwritten character input is sent to the first handwritten character recognition unit, and is interpreted as a recognition parameter change command “high-speed processing”. Then, the recognition speed is temporarily set to “high speed”, and the second handwritten character recognition unit recognizes the previous handwritten character information again.

  If the second handwritten character recognition cannot recognize the previous handwritten character information as “high-speed processing”, it can be determined that the handwritten character input by the user is not successful in the high-speed processing. That is, even for handwritten characters that the user will input in the future, it is expected that inconvenience will be caused by high-speed processing, so that the user is notified as warning information (405).

  Thus, the pattern recognition apparatus of the present invention can be applied to online handwritten character recognition.

  FIGS. 5A and 5B show an example in which the pattern recognition apparatus of the present invention is mounted on a copying machine.

  The pattern recognition performed by this apparatus is offline handwritten character recognition, and receives offline handwritten information as user input information. It is assumed that the user operates the copying machine by handwriting user setting information (502) for operating the copying machine on the setting sheet (501) and reading it from the document table of the copying machine.

  For example, as shown in FIG. 5A, it is assumed that user input information having a content of “density correction + 40” is taken into the apparatus. The first handwritten character recognition unit in the apparatus interprets this handwritten character input as a command “make density correction +40”. Here, it is assumed that the handwritten character recognition process of the copying machine performs image density correction as preprocessing, and holds the density correction parameters as recognition parameters. Then, “Set density correction to +40” is determined as a recognition parameter change command. Then, as indicated by the user, the density correction value is temporarily increased by 40, and the second handwritten character recognition unit recognizes the previous handwritten character information again.

  If the second handwritten character recognition cannot recognize the previous user input as “density correction + 40”, it is an excessive setting to increase the density correction by 40 with respect to the user input. And also in the handwritten character which the user inputs from now on, there exists a possibility that recognition performance may fall by increasing density correction by 40. Therefore, the user is notified as warning information (505) in the information presentation area (504) of the operation panel (503) of the copying machine.

  Thus, the pattern recognition apparatus of the present invention can be applied to off-line handwritten character recognition.

  6A and 6B are examples in which the pattern recognition apparatus of the present invention is mounted on a camera server.

  The pattern recognition performed by this apparatus is gesture recognition, and a user's gesture is received as moving image information as user input information. The camera server (601) captures the gesture operation of the user (604) from the camera (602) and recognizes it as an operation command of the camera server.

  The user can input a recognition parameter change command for gesture recognition in addition to a general operation command. As in the other embodiments, when the user's recognition parameter change command has an excessive value, the warning information (605) can be notified from the screen (603) on the camera server.

  As described above, the pattern recognition apparatus of the present invention can also be applied to gesture recognition.

  An object of the present invention is to supply a storage medium recording a program code of software that realizes the functions of the above-described embodiments to a system or apparatus, and the computer (or CPU or MPU) of the system or apparatus stores the storage medium. Needless to say, this can also be achieved by reading and executing the program code stored in.

  In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention.

  As a storage medium for supplying the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a magnetic tape, a nonvolatile memory card, a ROM, or the like can be used.

  Further, by executing the program code read by the computer, not only the functions of the above-described embodiments are realized, but also an OS (operating system) operating on the computer based on the instruction of the program code. It goes without saying that a case where the function of the above-described embodiment is realized by performing part or all of the actual processing and the processing is included.

  Further, after the program code read from the storage medium is written into a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, the function expansion is performed based on the instruction of the program code. It goes without saying that the CPU or the like provided in the board or the function expansion unit performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing.

3 is a functional configuration diagram of Embodiment 1. FIG. 2 is an operation flow according to the first embodiment. 6 is a functional configuration diagram of Embodiment 2. FIG. (A) It is an example of an outline of the portable information terminal of Example 3. (B) It is a warning output example of the portable information terminal of Example 3. (A) This is an example of setting paper to be read by the copying machine of the fourth embodiment. (B) Example of operation panel and warning output of copying machine of embodiment 4. (A) It is an example of an external appearance of the camera server of Example 5. FIG. (B) It is an example of a warning output of the camera server of Example 5.

Claims (9)

User input information capturing means for capturing user input information;
User input information storage means for storing the user input information;
Pattern recognition parameter storage means for storing pattern recognition parameters for determining the pattern recognition operation;
A first pattern recognition means for recognizing the user input information using the pattern recognition parameter and converting it to a first command;
Output means for outputting the first command;
When the first command is a command for changing the pattern recognition parameter, using the pattern recognition parameter according to the command, the user input information stored in the user input information storage means is recognized again, and the first command Second pattern recognition means for converting to a second command,
A pattern recognizing device comprising a comparing means for comparing the first command and the second command.   When the comparing means determines that the first command and the second command are different, it is considered that the user input pattern recognition parameter change command is an excessive setting, and presenting means for presenting an error or warning to the user The pattern recognition device according to claim 1, further comprising:   When the comparison means determines that the first command and the second command are the same, it is assumed that the user-recognized pattern recognition parameter change command is not an excessive setting, and the change to change the pattern recognition parameter The pattern recognition apparatus according to claim 1, further comprising means.   The pattern recognition apparatus according to claim 1, wherein the user input information is voice input information, and the pattern recognition process is a voice recognition process.   The pattern recognition apparatus according to claim 1, wherein the user input information is handwritten character input information, and the pattern recognition process is a handwritten character recognition process.   2. The pattern recognition apparatus according to claim 1, wherein the user input information is image input information, and the pattern recognition process is an image recognition process.   The pattern recognition apparatus according to claim 1, wherein the user input information is gesture input information, and the pattern recognition process is a gesture recognition process. User input information capturing step for capturing user input information;
A user input information storage step for storing the user input information;
A pattern recognition parameter storage step storing pattern recognition parameters for determining pattern recognition operation;
A first pattern recognition step of recognizing the user input information using the pattern recognition parameter and converting it to a first command;
An output process for outputting the first command;
When the first command is a command for changing the pattern recognition parameter, using the pattern recognition parameter according to the command, the user input information stored in the user input information storage means is recognized again, and the first command A second pattern recognition process to convert to a second command,
A pattern recognition method comprising a comparison step of comparing a first command and a second command.   A control program for causing a computer to execute the pattern recognition method according to claim 8.

Images