JP2017204067A - Sign language conversation support system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sign language conversation support system that allows an indication of intention to be accurately interpreted and allows the interpreted indication of intention to be accurately transmitted.SOLUTION: A sign language conversation support system includes: a sign language image reception device that receives a sign language image transmitted from a terminal device shooting the sign language image; a sign language image recognition device that recognizes the sign language image received by the sign language image reception device to generate sign language image data; a sign language image interpretation device that reads and interprets data from a sign language image storage device so that the sign language image data generated by the sign language image recognition device corresponds to a natural language; a natural language data generation device that generates the natural language from the sign language image data interpreted by the sign language image interpretation device; an audiovisual data generation device that generates the natural language generated by the natural language data generation device into audio/visible data; and an audio visual data transmission device that transmits the audiovisual data generated by the audiovisual data generation device.SELECTED DRAWING: Figure 5

Description

  The present invention relates to an apparatus, a system, and a computer program that support machine learning using artificial intelligence and communication performed using computer vision.

  Sign language is used as a means of communicating with hearing impaired people. For good communication between deaf people or between deaf and healthy people, both sides must understand sign language. However, it takes time and effort to understand and use sign language freely, and it cannot be said that it is popular in a wide range of age groups. Therefore, in order for the sign language to accurately convey the intention of the sign language to an unspecified number of people, the sign language is converted into image data by the image acquisition means, and the sign language confirmation method for recognizing the converted image data as characters and An apparatus is disclosed (for example, refer to Patent Document 1).

JP 2012-252581 A

  However, as there are dialects of words and personality in speaking, hand and finger movements vary from person to person when speaking sign language. Therefore, even if it is attempted to acquire sign language image data and convert it into character data, there is a problem that the intention representation cannot be accurately transmitted without being accurately translated. On the other hand, when generating sign language video based on text information or voice information, it is possible to increase the familiarity of communication if the individuality of the speaker can be expressed, but in simple conversion work, it is a uniform sign language The reality is that only videos can be generated. Further, the method of converting a sign language image into characters and voices has a problem that accurate translation cannot be performed unless an image is taken from the front of the speaker in sign language. One embodiment of the present invention has an object to solve such a problem.

  A sign language conversation support system according to an embodiment of the present invention includes a sign language image recognition device that recognizes a sign language image, and a sign language image conversion unit that translates the sign language image recognized by the sign language image recognition device into a natural language. The translation apparatus is configured to perform matching with sign language image data previously machine-learned by machine learning.

  A sign language conversation support system according to an embodiment of the present invention recognizes a sign language image receiving device that receives a sign language image transmitted from a terminal device that captures a sign language moving image, and a sign language image received by the sign language image receiving device. A sign language image recognition device that generates image data, a sign language image translation device that reads and translates data from a sign language image storage device so that the sign language image data recognized by the sign language image recognition device corresponds to a natural language, and sign language image recognition Natural language data generating device for generating natural language from sign language image data translated by the device, audiovisual data generating device for generating natural language generated by the natural language data generating device into audiovisual data, and audiovisual data generating device And an audiovisual data transmission device that transmits the audiovisual data generated in (1).

It is a figure which shows the outline | summary of the sign language conversation assistance system which concerns on one Embodiment of this invention. It is a figure which shows the whole structure of the sign language conversation assistance system which concerns on one Embodiment of this invention. It is a figure which shows the whole structure of the sign language conversation assistance system which concerns on one Embodiment of this invention. It is a figure which shows the structural example of the hardware of the computer apparatus 102 contained in the sign language conversation assistance system which concerns on one Embodiment of this invention. It is a figure which shows the functional structure of the sign language conversation assistance system which concerns on one Embodiment of this invention. It is a figure which shows an example of the image process performed in the sign language conversation assistance system which concerns on one Embodiment of this invention. It is a figure explaining one form which image | photographs the sign language performed in the sign language conversation assistance system which concerns on one Embodiment of this invention. It is a figure which shows the functional structure of the sign language conversation assistance system which concerns on one Embodiment of this invention. It is a figure which shows the functional structure of the sign language conversation assistance system which concerns on one Embodiment of this invention.

  Embodiments of the present invention will be described with reference to the drawings. However, the present invention can be implemented in many different modes and should not be construed as being limited to the description of the embodiments exemplified below. In order to clarify the description, the drawings may be schematically represented with respect to the width, thickness, shape, and the like of each part as compared to actual aspects, but are merely examples and limit the interpretation of the present invention. It is not a thing. In addition, in the present specification and each drawing, the same elements as those described above with reference to the previous drawings are denoted by the same reference numerals (or reference numerals with a, b, etc. added to the numerals), and detailed description will be given. It may be omitted as appropriate. In addition, the letters “first” and “second” attached to each element are convenient signs used to distinguish each element, and have no meaning unless otherwise specified. .

  A sign language conversation support system according to an embodiment of the present invention has a function of performing a process of recognizing a sign language image and converting it into a natural language, and a process of recognizing the natural language and generating a sign language image. This sign language conversation support system is connected to a telecommunication line (including a telephone line and an Internet line) and operates in a state where it can communicate with a user terminal device.

  FIG. 1 shows an outline of a sign language conversation support system. The sign language conversation support system 100 includes a computer device 102 and is connected to a terminal device 104 on the user side via an electric communication line. The operation of the sign language uttered by the user is captured by the terminal device 104. The terminal device 104 is various electronic devices capable of capturing an image, and a video camera, a camera capable of capturing an operation, and the like are applied in addition to a video camera. The sign language image captured by the terminal device 104 is transmitted to the computer device 102 as a moving image or as a plurality of continuous still images. The terminal device 104 and the computer device 102 are connected by an electric communication line. Hereinafter, the sign language image includes not only a moving image but also one or a plurality of still images.

  The sign language image transmitted from the terminal device 104 is translated into a natural language by a computer. The computer stores sign language images. The computer apparatus 102 refers to the transmitted sign language image and a pre-stored sign language image, and translates into a natural language.

  The computer apparatus 102 uses a sign language image captured by the user as a learning target. The computer apparatus 102 performs machine learning on the sign language image of the user by deep learning. Thereby, the sign language images of many users are accumulated, and the sign language image analysis accuracy is improved by deep learning. That is, the user utters sign language according to common rules, but each person's sign language action includes personality. In other words, there is a fluctuation in the sign language action, so that the accuracy of translation is reduced in the image analysis of the uniform sign language action. However, by using sign language images collected from users for deep learning, it becomes possible to increase the discriminating power of sign language images and improve the translation system.

  The computer apparatus 102 obtains a sign language image transmitted from the user and matches the learned sign language image in real time, and outputs (transmits) the result to the user or another user.

  FIG. 2 shows the overall configuration of the sign language conversation support system 100. In the sign language conversation support system 100, a computer device 102, a terminal device 104 of one conversation person, and a terminal device 106 of the other conversation person are connected via an electric communication line. As the terminal devices 104 and 106, mobile phones, tablet terminals, multifunctional mobile phones called smartphones, and the like are applied. In addition, as a communication means belonging to the conversation person, an electronic device that can communicate with each other through an electric communication line such as a personal computer 108 is applied. The conversation content (audio / visual data in sign language and natural language) may be an electronic device that can mainly output only information, such as a sound reproduction device 110 such as a radio and a television 112.

  The sign language conversation support system according to the present embodiment is capable of two-way conversation via sign language between a plurality of terminal devices. For example, as shown in FIG. 3, when a sign language conversation is transmitted from a certain terminal device 116, the computer device 102 translates the sign language and transmits the conversation to another terminal device 114. For example, the terminal device 114 is translated sign language to natural language "Hello, how are you?" Is output to the voice, by the character and / or animation, or the like. Next, the user of the terminal device 114 is "Hello. Yes, I'm fine." If you reply with a letter or voice and, computing device 102 to translate this into sign language, to send as sign language image to the terminal device 116.

  The sign language conversation support system according to the present embodiment can execute such a conversation in real time. This also makes it possible to make a call by sign language. Conversely, information can be transmitted from another terminal device to the sign language terminal device. If the information is voice or text, the information can be converted into a sign language video. This enables two-way remote communication.

  FIG. 3 shows a one-to-one relationship between a terminal device of a sign language speaker and a terminal device of a speaker such as a voice, but the embodiment of the present invention is not limited to this. For example, sign language transmitted from a terminal device of a sign language speaker may be transmitted to a plurality of terminal devices in a form translated into a natural language. In addition, a sign language image may be transmitted from one terminal device that utters text or voice to a plurality of terminal devices.

  FIG. 4 shows a hardware configuration example of the computer apparatus 102 included in the sign language conversation support system according to the embodiment of the present invention. The computer device 102 includes a central processing unit 120 that performs arithmetic processing and various controls, an image processing device 122 that analyzes and generates image data, a main storage device 124 that stores data, programs, and programs including deep learning algorithms, It includes a sign language image storage device 118 that stores data such as sign language images, a receiving device that receives data and the like via a telecommunication line, and a transmitting device 128 that transmits data and the like to the telecommunication line.

  The central processing unit 120 is realized by a microprocessor, and the image processing apparatus is realized by a dedicated processor specialized for image processing. The main storage device 124 is realized by a semiconductor memory such as a dynamic random access memory (DRAM). The sign language image storage device 118 is composed of a rewritable nonvolatile memory, and for example, a nonvolatile semiconductor memory called a magnetic disk (hard disk) or flash memory is applied. The receiving device 126 and the transmitting device 128 are realized by communication modules conforming to various standards. The computer device 102 realizes a sign language conversation support function by connecting these devices in common via a bus line and operating in cooperation.

  FIG. 5 shows a functional configuration of a sign language conversation support system according to an embodiment of the present invention. The sign language conversation support system 100 includes a sign language image reception device 130, a sign language image recognition device 132, a sign language image translation device 134, a natural language data generation device 136, a natural language data analysis device 138, a natural language calibration device 140, and a sign language image learning device 142. A sign language image data storage device 154 including a sign language image data storage device 148, a sign language image data storage device 148, a standard sign language image data storage device 150, and a learning sign language image data storage device 152, and a natural language dictionary storage device 156. Including.

  The sign language image receiving device 130 receives the sign language image transmitted from the user terminal device 104. The sign language image received by the sign language image reception device 130 is output to the sign language image recognition device 132. The sign language image recognition device 132 stores the received sign language image in the sign language image storage device 148. The sign language image recognition device 132 extracts a characteristic part of the sign language image. For example, the sign language image recognition device 132 recognizes a position where the signer's head and the hand are not relative to each other. Moreover, in order to recognize the form of the hand which performs sign language, the outline part of a signer's hand is extracted and an outline is produced | generated. As described above, the sign language image recognition device 132 performs morphological image processing. The sign language image recognition device 132 generates sign language image data including a signer image and image data subjected to morphological image processing as a set. Alternatively, the sign language image recognition device 132 may generate image data that has undergone morphological image processing as sign language image data.

  For example, when the sign language image shown in FIG. 6A is input to the sign language image recognition device 132, the sign language image is image-processed as shown in FIG. Superimpose lines or images with recognized shapes. When sign language image recognition or machine learning is performed, it is performed based on the sign language image subjected to the morphological image processing. Thereby, the sign language operation can be simplified and clarified, and the recognition accuracy of the sign language image in the image processing can be improved. In machine learning, learning accuracy and proficiency can be improved. However, such image processing is not essential, and learning and real-time processing may be performed without performing image processing.

  In FIG. 5, the sign language image recognition device 132 outputs the generated sign language image data to the sign language image translation device 134. The sign language image storage device 148 refers to the standard sign language image data stored in the standard sign language image data storage unit 152 of the sign language image data storage device 154, and some of the standard sign language image data matches the sign language image data. Check whether or not. In addition, the sign language image translation device 134 refers to the learning sign language image data stored in the learning sign language image data storage unit 152 and checks whether there is a match with the sign language image data among the learning sign language image data. . The sign language image translation device 134 associates sign language image data with standard sign language image data and / or learning sign language image data.

  There are dialects in sign language. For example, sign language in an area in Kansai is often expressed with large gestures using both hands. For example, when expressing “No way!”, The palm of both hands is moved in front of the chest, but in other areas, the fist is applied to the chest with one hand and then the palm is opened and pushed out. Therefore, when only standard sign language image data is referred to, there are cases where sign language cannot be correctly translated into a natural language. However, in the sign language conversation support system according to the present embodiment, learning data of sign language images deep-learned with artificial intelligence is stored in the learning sign language image data storage unit 152. The sign language image translation device 134 refers to the data in the learning sign language image data storage unit 152 and performs an operation of applying to which area dialect the sign language image is close or to which area dialect.

  The sign language image generation device 136 converts the sign language image data associated with the standard sign language image data and / or the learning sign language image data by the sign language image translation device 134 into a natural language. For example, if the standard sign language image data and / or learning sign language image data with sign language image data associated with it means "Hello", and generates a natural language data of "Hello" as a natural language.

  The natural language data generated by the natural language data generation device 136 is output to the natural language data analysis device 138. The natural language data analysis device 138 analyzes the natural language data and refers to the natural language dictionary storage device 156 to correct grammatical errors and vocabulary into natural Japanese. In addition, when the sign language speaker speaks in a dialect, the natural language data analysis device 138 corrects the language according to the dialect. For example, when the sign language operation, which means "hello" has been translated is translated by sign language image translation device 134 as "Hello", if the speaker is using the Okinawan dialect, the "Hello", "high side To the word "." Thus, by generating or correcting natural language data according to the dialect of the sign language speaker, sign language translation that respects the individuality of the speaker can be performed.

  The natural language data corrected by the natural language data analysis device 138 is output to the natural language calibration device 140. The natural language calibration device 140 identifies the sign language image data that is the object of calibration, and outputs the information to the sign language image learning device 142. The sign language image learning device 142 reads out the corresponding sign language image from the sign language image storage device 148 and adds it to the data in the learning sign language image data storage unit 150 in association with the corrected natural language. Thereby, sign language image data is accumulated in the learning sign language image data storage unit 152.

  By storing the natural language information corresponding to the sign language image data in the learning sign language image data storage unit 152, it is possible to improve the accuracy of sign language translation even if there is a fluctuation in the behavior of the sign language. Moreover, even if the action of the sign language is based on a dialect, it is possible to translate the sign language according to the dialect.

  The audiovisual data generation device 144 converts the natural language data generated by the natural language data analysis device 138 into audiovisual data. For example, the audiovisual data generation device 144 converts natural language data into audio data, character data, and / or image data representing emotions and expressions. The audiovisual data transmission device 146 outputs the audio data, character data, and / or image data such as emotion and expression generated by the audiovisual data generation device 144 to the terminal device 106 of another user.

  As described above, the sign language conversation support system according to the present embodiment can realize communication reflecting the personality of the sign language speaker. In this case, the sign language conversation support system can improve the accuracy of sign language translation by learning the sign language image by deep learning, thereby recognizing variations in the movement of the speaker and dialects. Furthermore, the sign language conversation support system according to the present embodiment can execute communication via such sign language in real time.

  As shown in FIG. 7, the sign language speaker is the camera function of the video camera or portable terminal device 104 worn, the camera (imaging device) 158 installed behind the camera, the drone 160 equipped with the camera, and the other behind the sign language speaker. It is also possible to use a device for taking pictures from. In this way, by photographing the sign language from an angle other than the front, the sign language speaker can speak the sign language in a relaxed state without being aware of the camera.

  However, when shooting from behind a sign language speaker, the video shot by the camera is different from the video shot from the front. Images taken from behind a sign language speaker are basically in the opposite direction, and differ depending on the angle at which they are taken.

However, in the present embodiment, in order to correct such a problem and accurately recognize the sign language image, one or a plurality of combinations of the following measures are performed.
(1) Machine learning of sign language images from behind at various angles and distances in advance.
(2) Before performing machine learning, video processing that is close to a plurality of angles is performed on a sign language image that is a learning material photographed from the front or from behind, and all of them are subjected to learning.
(3) The sign language image input at the time of real time processing is adjusted and converted in real time, and the recognition processing is performed in a form close to the machine-learned front or back sign language image.

  As described above, the sign language conversation support system according to the present embodiment recognizes the sign language image accurately by learning the sign language image captured from behind the sign language speaker using the artificial intelligence function. be able to.

  FIG. 8 shows an example of a sign language conversation support system 100b that can be adapted even when a sign language image is taken from an angle other than the front of the sign language. The sign language conversation support system 100b shown in FIG. 8 is different from the sign language conversation support system shown in FIG. 5 in that it includes a sign language image analysis device 162, a three-dimensional image storage device 164, and a three-dimensional image generation device 166. Hereinafter, the difference will be mainly described.

  The sign language image analysis device 162 analyzes from which angle the captured sign language image is captured. The three-dimensional image storage device 164 stores sign language images taken in advance from various angles and distances from the front and angles other than the front. The sign language image analysis device 162 reads the data of the sign language image from the three-dimensional image storage device 164 and identifies from which angle the captured sign language image is an image. Note that the photographed sign language image may include information photographed from a plurality of angles. By using sign language images taken from a plurality of angles, it is possible to improve the analysis accuracy of sign language images.

  Machine learning may be performed in the sign language image analysis apparatus 162. In the machine learning, it is preferable to perform video processing close to a plurality of angles on a sign language image photographed from the front and the back, and make all of them subject to learning. In addition, when a conversation through sign language is performed in real time, the sign language image is adjusted and converted in real time, and the recognition process is performed in a form close to the machine-learned front or back sign language image. It is preferable to do.

  The sign language image analysis device 162 outputs sign language image data to the three-dimensional image generation device 166 together with the specified shooting angle information. The three-dimensional image generation device 166 reads out the three-dimensional sign language image data photographed from various angles corresponding to the operation of the sign language from the three-dimensional image storage device 164 based on the specified photographing angle information and the sign language image. Combine to generate a 3D image of sign language. The 3D image generation device 166 outputs the generated 3D sign language image data to the sign language image recognition device 132. The sign language image recognition device 132 recognizes 3D sign language image data. The subsequent operations and processes of the respective units are the same as the processes shown in FIG.

  Further, the sign language image recognition device 132 may recognize the contents of the sign language based on the specified shooting angle information and the sign language image without using the 3D image generation device 166. The sign language image translation device 134 performs recognition processing in a form close to a machine-learned front or back sign language image, and performs sign language translation.

  Thus, according to the present embodiment, even when the sign language image is taken from an angle other than the front, the sign language translation can be performed accurately, and between the sign language speaker and the natural language speaker. Smooth communication can be achieved.

  FIG. 9 shows an example of a sign language conversation support system that converts natural language speech into sign language. The sign language conversation support system 100c includes an audiovisual data reception device 168, an audiovisual data conversion device 170, a natural language data analysis device 138, a sign language image conversion data generation device 172, a sign language image translation device 134, a sign language image conversion data generation device 172, and a sign language image. It includes a transmission device 174, a sign language image data storage device 154 including a standard sign language image data storage unit 150 and a learning sign language image data storage unit 152, and a natural language dictionary storage device 156. Such a sign language conversation support system 100c is realized by the computer apparatus 102 shown in FIG.

  The audiovisual data receiving device 168 receives audiovisual data such as voice, characters, and animation transmitted from the user terminal device 106. The audiovisual data received by the audiovisual data receiving device 168 is output to the audiovisual data conversion device 170. The audiovisual data converter 170 converts audiovisual data such as voice, characters, animation, etc. into data that can be recognized by a computer (for example, binary format data).

  The audiovisual data converted by the audiovisual data converter 170 is output to the natural language data analyzer 138. The natural language data analysis device 138 analyzes audiovisual data and refers to the natural language dictionary storage device 156 to convert grammatical errors and vocabulary into sentences that can be translated into sign language. In addition, when the audiovisual data includes a dialect, the natural language data analysis apparatus 138 corrects the dialect according to the dialect.

  The sign language image conversion data generation device 172 divides a natural language sentence into one or a plurality of blocks so that it can be translated into sign language. For example, "Hello, I am it is hot today" natural language is a case was a sentence, as "Hello", and "today", and "hot", broken down into blocks of the "This is it".

  The sign language image translation device 134 reads the sign language image data from the sign language image storage device 148, combines the sign language images corresponding to the natural language, and generates one set of sign language image data. In the above example, "Hello", "today", "hot", combined with the sign language images corresponding to the "This is it". The sign language image translation device 134 generates these sign language images as one continuous sign language image.

  In this case, when the sign language image translation device 134 reads data from the learning sign language image data storage unit 152, the sign language image translation device 134 may read data based on the personality of the natural language speaker. For example, when sign language image data of a natural language speaker is stored, the data of the corresponding person may be read out. In addition, when the utterance in the natural language is a dialect, the sign language image translation device 134 may read a sign language image corresponding to the dialect. Thus, by making the personality of the natural language speaker reflect, it is possible to increase the closeness between the sign language and the natural language conversation. As described with reference to FIG. 5, the learning sign language image data storage unit 152 stores data as needed by machine learning. Such learning data for sign language images can also be used when sign language image data is generated from audiovisual data in natural language. The sign language image conversion data generation device 172 generates a sign language image by animation, for example. The sign language image transmission device 174 transmits the sign language image to the terminal device 104 of the user.

  According to this embodiment, information based on a natural language can be converted into a sign language image. In this case, the sign language image can be generated in accordance with the personality of the speaker because the image data of the sign language is machine-learned.

  5, 8, and 9 have been described as the functional configuration of the computer apparatus 102, one embodiment of the present invention is not limited to this, and such a function is used in a user terminal device. It can also be realized.

DESCRIPTION OF SYMBOLS 100 ... Sign language conversation support system, 102 ... Computer apparatus, 104 ... Terminal apparatus, 106 ... Terminal apparatus, 108 ... Personal computer, 110 ... Sound reproduction apparatus, 112 ... Television John, 114 ... terminal device, 116 ... terminal device, 118 ... sign language image storage device, 120 ... central processing device, 122 ... image processing device, 124 ... main storage device, 126 ... Receiving device, 128 ... Transmitting device, 130 ... Sign language image receiving device, 132 ... Sign language image recognition device, 134 ... Sign language image translation device, 136 ... Natural language data generation device, 138: Natural language data analysis device, 140: Natural language calibration device, 142: Sign language image learning device, 144: Audiovisual data generation device, 146: Audiovisual data Data transmission device, 148 ... sign language image storage device, 150 ... standard sign language image data storage unit, 152 ... learning sign language image data storage unit, 154 ... sign language image data storage device, 156 ... natural Language dictionary storage device, 158 ... camera, 160 ... drone, 162 ... sign language image analysis device, 164 ... three-dimensional image storage device, 166 ... three-dimensional image generation device, 168 ... Audiovisual data receiving device, 170 ... audiovisual data conversion device, 172 ... sign language image conversion data generation device, 174 ... sign language image transmission device

Claims (12)

A sign language image recognition device for recognizing a sign language image, and a sign language image translation device for translating the sign language image recognized by the sign language image recognition device into a natural language,
The sign language image translation device performs matching with sign language image data machine-learned in advance by machine learning.   The sign language conversation support system according to claim 1, wherein the sign language image translation device performs matching with sign language image data previously machine-learned in real time.   The sign language conversation support system according to claim 1, wherein the sign language image recognition device performs morphological image processing on an image of a sign language speaker.   4. The sign language conversation support system according to claim 3, wherein the morphological image processing extracts a position of a head of a sign language speaker and a contour line of the hand. The sign language image is an image taken from an angle other than the front,
The sign language image recognition device determines an angle taken from the sign language image;
2. The sign language conversation support system according to claim 1, wherein the sign language conversation translation device performs matching with the sign language image data based on sign language image data that is photographed in advance from a plurality of angles and distances and machine-learned. A sign language image receiving device that receives a sign language image transmitted from a terminal device that captures a sign language image;
A sign language image recognition device that recognizes a sign language image received by the sign language image reception device and generates sign language image data;
A sign language image translation device that reads and translates data from a sign language image storage device so that the sign language image data recognized by the sign language image recognition device corresponds to a natural language;
A natural language data generation device for generating a natural language from the sign language image data translated by the sign language image recognition device;
An audiovisual data generation device for generating natural language data generated by the natural language data generation device into audiovisual data;
A sign language conversation support system, comprising: an audiovisual data transmission device that transmits the audiovisual data generated by the audiovisual data generation device.   The sign language conversation support system according to claim 6, wherein the sign language image translation device performs matching with sign language image data previously machine-learned in real time.   The sign language conversation support system according to claim 6, wherein the sign language image recognition device performs morphological image processing on the sign language image.   The sign language conversation support system according to claim 8, wherein the morphological image processing extracts a head position and a hand outline of a sign language speaker.   The sign language conversation support system according to claim 6, wherein the sign language image is an image photographed from an angle other than the front, and further includes a sign language image analysis device that analyzes the sign language image. The sign language image recognition device determines an angle taken from the sign language image;
11. The sign language conversation support system according to claim 10, wherein the sign language conversation translation device performs matching with the sign language image data based on sign language image data that has been photographed from a plurality of angles and distances and machine-learned in advance. The sign language image storage device includes a standard sign language image storage unit that stores standard sign language images, and a learning sign language image storage unit that stores sign language images learned by machine learning,
The sign language conversation support system according to claim 6, wherein the sign language image translation device reads data from the standard sign language image storage unit and the learning sign language image storage unit.