JPH07311545A - Finger language interpretation device - Google Patents

Abstract

PURPOSE:To provide a finger language interpretation device which enables a user to input finger language in a finger language sentence unit and to recog nize the finger language with good accuracy even if person to input the finger language changes or posture to make the finger language changes. CONSTITUTION:The time of the boundary of finger language words composing a finger language sentence is detected in a boundary detector 3. This boundary detector 3 sends a detection signal d2 to a finger language data memory device 4 when the detector detects the boundary of the finger language words. The normalized finger language data sent from a normalizing device 2 is previously stored in the finger language data memory device 4 which sends the finger language data stored thus far to a finger language recognizer 5 when the detection signal d2 is set to the memory device. What finger language is the inputted finger language is recognized in this finger language recognizer 5 by using the finger language data of every finger language word inputted from the finger language memory device 4 and the finger language word data stored in a finger language word dictionary 6. The recognized results are sent to an output device 7 and are outputted in the form of characters and voices.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sign language interpreter for recognizing a sign language, that is, a gesture as a sign language, and translating a sign language sentence consisting of connection of sign language words into a Japanese sentence. The present invention relates to a sign language interpreting device that accurately detects a boundary and normalizes it so that a sign language sentence can be converted into a Japanese sentence even when an unspecified person performs sign language.

[0002]

2. Description of the Related Art In sign language, a plurality of sign language words are continuously connected to form a conversation composed of one sentence. The sign language interpreter is a device that translates and outputs a sign language sentence composed of such continuous sign language words into a Japanese sentence. In a conventional sign language interpreter, a method of inputting and recognizing a sign language word is generally used. That is, when one sign language word is input, this is translated, one Japanese word corresponding to this is output, and by repeating this, the sign language sentence is translated. Among them are
There are sign language recognition techniques in which sign language words in a normal sign language conversation sentence are continuously expressed, but these are only to the extent that experiments are performed only on a small number of sentences. Sign language data used for sign language recognition is a sign language input device, for example, the sign language data input from a sign language input device consisting of a detector called a data glove that detects a hand gesture for each part such as a joint. It is used as sign language data. In addition, what grasped the sign language as the facial expression recognition method of the facial image recognition device is, for example, Japanese Patent Application No. 4-24728.
There is a specification and a drawing. Further, one that recognizes an input sign language word by using the sign data stored in the word dictionary is, for example, Japanese Patent Application No. 4-2356.
No. 33 and drawings.

[0003]

In order to interpret general sign language, it is necessary to recognize a sign language sentence expressed by a sequence of several sign language words. However, in conventional sign language recognition, by inputting each sign language word,
Since a method of recognizing each sign language word is usually used, there is a problem that a general sign language interpreter, that is, a sign language sentence continuously input at high speed cannot be translated. On the other hand, there are attempts to recognize continuous sign language sentences, but when sign language words are continuously expressed, there is a problem that the sign language data changes largely around the boundaries of the sign language words and the recognition rate decreases.
There is also a problem that it is not easy to distinguish the position of the sign language word, that is, the position of the hand gesturing the word from the boundary of the sign language word. Further, the sign language data input from the sign language input device changes variously depending on the physique of the person who performs the sign language, the position and the direction in which the sign language is performed. In the conventional sign language recognition method, the sign language data input from the sign language input device is used as it is for recognition. Therefore, if the posture of the person who performs sign language changes from the middle, or if a taller person or a shorter person performs sign language, the recognition rate will drop significantly, and only the person who has registered the sign language data in the system in advance. Then, there was a problem that a high recognition rate could not be obtained. A first object of the present invention is to solve these conventional problems and to provide a sign language interpreter capable of inputting sign language in units of sign language sentences and recognizing sign language sentences with high accuracy. A second object of the present invention is to provide a sign language interpreting device that can accurately recognize sign language even if the person who inputs the sign language changes or the posture of the person who performs the sign language changes.

[0004]

In order to achieve the above object, the sign language interpreting apparatus of the present invention specifies a sign language word boundary in order to accurately recognize sign language words continuously expressed in a sign language sentence. By clarifying the boundaries of the sign language words by inserting gestures and facial expressions, or by inputting a signal from another input device at the boundaries of the sign language words, recognition of sign language sentences can be performed by recognition of each sign language word. . In addition, in order to absorb changes in the posture in which sign language is input and differences in the physique of the person who performs sign language, by entering a specific position and direction in advance before entering sign language, new sign language data can be entered. Normalize.

[0005]

In the present invention, since the sign language sentence can be accurately divided into the sign language words before recognition, it is possible to reduce the misrecognition due to the sign language word boundary, and to recognize the sign language sentence with high accuracy into a Japanese sentence It will be possible to convert. In addition, by normalizing the sign language data before inputting the sign language, it is possible to absorb changes in the sign language data due to the physique of the person who performs the sign language and the posture of performing the sign language, so it is possible to In addition, it is possible to accurately recognize sign language words and sign language sentences and translate them into Japanese sentences.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings. FIG. 1 is an overall block diagram of a sign language interpreter showing an embodiment of the present invention. In FIG. 1, reference numeral 1 denotes a sign language input device for inputting sign language, for example, an input device such as a glove (data glove) (trademark) that detects a hand gesture for each joint and inputs coordinate and angle data. A normalization device for normalizing the input sign language data, for example, a memory storing a table that records the physique, position, etc. of the person who performed the sign language, 3 detects the boundaries of the sign language words from the input sign language data A boundary detection device, such as a means for detecting a stop of a hand, a means for detecting a facial expression indicating a boundary, or a means for detecting a hand gesture indicating a boundary, 4 is a sign language data storage device for storing sign language data for each sign language word. For example, the device 5 which receives the information d2 from the boundary detection device 3 and divides the continuously input sign language data into sign language words and sends it to the next sign language recognition device 5 is the input sign language data? A sign language recognition device for recognizing a sign language word, for example, a device for determining a matching word by comparing and referring to the same word in the sign language word dictionary, 6 is a sign language word dictionary used for recognition For example, as shown in FIG. 3, which will be described later, the position coordinates of a person who performs sign language for normalization, data such as coordinates and angles at which the gesture of the sign language word is detected for each joint, and the name of the corresponding sign language word are shown. The stored word dictionary, 7 is an output device for outputting the recognized sign language words, for example, a display device,
It is a speaker or a printer.

FIG. 2 is a diagram showing the data structure of the sign language data d1 input from the sign language input device in FIG.
The sign language input device 1 is well known as a glove (data glove) having a device for converting the bending state of each finger of the hand, the position and direction of the hand into an electric signal as described above. The data structure has a hand position x, for each of the right and left hands.
It is composed of y, z (coordinates), hand directions α, β, γ (vectors), and finger bending angles α1, β1 to α5, β5 (angles). Reference numerals 21, 22, and 23 indicate storage locations (addresses) of data at sampling times T0, T1, and Tn, and 24 and 25 indicate storage locations of right-hand data and left-hand data, respectively. 2411 is data representing the x-coordinate position of the right hand, 2412 is data representing the y-coordinate position of the right hand, and 2413 is data representing the z-coordinate position of the right hand. 2421 is a vector perpendicular to the palm of the right hand (see FIG.
24) is data representing the y-coordinate component of the vector perpendicular to the palm of the right hand, 2423 is the data representing the z-coordinate component of the vector perpendicular to the palm of the right hand, 2431 Is data representing the x-coordinate component of a vector (see 62 in FIG. 6 described later) parallel to the palm of the right hand, and 2432 is data representing the y-coordinate component of the vector in the fingertip direction parallel to the palm of the right hand, 2433 is data representing the z-coordinate component of the vector in the fingertip direction parallel to the palm of the right hand, and 2441 is the x-coordinate of the vector parallel to the palm of the right hand and perpendicular to the side surface on the thumb side (see 63 in FIG. 6 described later). Data representing the component, 2442 is data representing the component of the y coordinate of a vector parallel to the palm of the right hand and perpendicular to the side surface on the thumb side, 244.
Data 3 represents the z-coordinate component of a vector parallel to the palm of the right hand and perpendicular to the side surface on the thumb side. 2451 is data representing the angle of the first joint of the first finger of the right hand (joint in the center of the fist), and 2452 is data representing the angle of the second joint of the first finger of the right hand (joint in the center of each finger). , 2453 is data representing the angle of the first joint of the fifth finger of the right hand, and 2454 is data representing the angle of the second joint of the first finger of the right hand.

FIG. 3 is a diagram showing a data structure of a sign language word dictionary used for recognition. The data recorded in the sign language word dictionary 6 are the sign language data read from the sign language input device 1 for each sign language word, and the data used for normalizing the sign language data. In FIG. 3, 31 is the position of the parietal region that serves as a reference for normalization, that is, the position coordinates of the parietal region of the person who performed the sign language, 32 is the position of the right shoulder that is the reference for normalization, and 33 is Left shoulder position, which is the reference for 3
4 is the position of the navel which serves as a reference for normalization (see FIG.
44), and 35 is the position of the hand when the arm serving as a reference for normalization is extended straight forward (45 in FIG. 4, which will be described later).
), 36 is data of sign language words. 361, 36
2, 363 is data for each sign language word, and the inside is shown in FIG.
Similar to the data structure of, the hand position (coordinates), hand direction (vector), finger bending angle (angle), and corresponding sign language word names are stored for each right hand and left hand.

The operation of the sign language interpreter of FIG. 1 will be described. FIG. 4 is a diagram showing reference points input to normalize the sign language position and physique. The sign language data input from the sign language input device 1 is input to the normalization device 2 and the position,
The direction and physique are normalized. The operation of the normalization device will be described with reference to FIGS. First, in order to normalize the physique and position, as shown in FIG. 4, the coordinates of the position of the crown 41, the right shoulder 42, the left shoulder 43, the navel 44, and the position of the hand 45 when the arm is straightened forward are input. To do. The position of the hand may be either the right hand or the left hand. Input the position of the crown (Xh, Yh, Zh) and the position of the right shoulder (Xsr,
Ysr, Zsr), the position of the left shoulder (Xsl, Ysl, Z
sl), the navel position (Xn, Yn, Zn), and the hand position when the arm is straightened forward (Xp, Yp, Z
p). Also, the positions of the hands recorded when the crown, both shoulders, navel and arm are extended forward are recorded in the sign language word dictionary 6. The position of the crown recorded in the sign language word dictionary 6 is (xh, yh, zh), and the position of the right shoulder is (xsr, ys).
r, zsr), the position of the left shoulder is (xsl, ysl, zs
l), the navel position is (xn, yn, zn), and the hand position when the arm is extended forward is (xp, yp, zp).
First, the position data in the sign language data is expressed as a relative position from the top of the head in order to absorb the deviation of the position where the sign language is performed. That is, if the position before conversion is (X0, Y0, Z0) and the position after conversion (X1, Y1, Z1), the position after conversion is expressed by (Equation 1).

[Equation 1]

FIG. 5 is a diagram showing the transformation of the coordinate system by normalizing the position, FIG. 6 is a diagram showing a vector representing the direction of the hand, and FIG. 7 is a view showing the hand in front of the body. It is a figure which shows the relationship between the direction of a hand and a coordinate system in the case of. By the conversion by the normalization, the position data represented by the coordinate systems 511 and 512 in FIG. 5 is represented by the coordinate systems 521 and 522. That is, the input position data (511, 512) is converted into a coordinate system with the position of the crown as a reference. Further, as the position reference, the navel position may be used instead of the crown. That is, it can be converted into a coordinate system with the position as a reference. Further, when the sign language is always input while sitting on a chair, the position of the hand when the hand is placed on the knee instead of the position of the navel can be used as a reference.

Next, normalization of the sign language is performed. When the coordinate systems 521 and 522 as shown in FIG. 5 are used, the conversion of the sign language direction is limited to the xy plane, so it is sufficient to input the reference direction on the xy plane. That is, as is apparent from FIG. 5, when the vector is viewed from the upward direction z, the operation is performed only in the x and y directions. For this reason, the directions of the left and right hands in a state where both hands are brought together on the front of the body are input, and the directions on the xy plane are obtained from them. As shown in FIG. 6, the data of the direction input from a well-known glove input device is used to calculate the vector 61 of the direction perpendicular to the palm, the vector 62 of the direction of the fingertips of the palm, and the lateral direction of the palm perpendicular to them. It can be converted into a vector 63. With the palms in front of you, as shown in Figure 7,
Palm tip direction 71 or palm side vector 72
The mapping of the xy plane of becomes large. Here, the map is a locus in the x and y directions seen from above. Looking at the loci in the x and y directions from the z direction, the angle in the y direction is larger than the angle in the x direction with respect to z, and thus the mapping in the x direction is large.
Therefore, first, of the palm fingertip direction 71 and the palm lateral direction vector 72, the vector having the larger mapping to the x axis (71 in FIG. 7) is selected, and the average vector of the right hand and left hand vectors is calculated. . Thereby, the direction of the front of the person can be obtained. That is, in the state in which both hands are put together, the palm and the front-back direction of the body are parallel to each other, so that the mapping of the average vector onto the xy plane represents the front direction of the body.

FIG. 8 is a diagram showing the relationship between the body direction and the coordinate system. As shown in FIG. 8, xy of this average vector 81
The x-axis component and the y-axis component of the mapping on the plane are respectively Xa8
2, and Ya83, the body direction Θ84 is given by (Equation 2).

[Equation 2] In order to absorb the change in the data depending on the orientation of the body, the coordinates of the position after the position normalization (X plane) (X
1, Y1) is converted into a position on the x ′ axis 861 and the y ′ axis 862, which is obtained by rotating the x axis 851 and the y axis 852 by Θ. This conversion operation is given by (Equation 3). Also, for three types of vectors representing the direction of the hand, the components of the vector in the x-axis, y-axis, and z-axis directions are given by
Can be normalized by performing the conversion of.
By this conversion, the sign language data is always converted into data in the coordinate system with the crown as the origin and the front direction of the body as the x-axis. In addition to normalizing the direction, in addition to using the direction when both hands are put together, place the hand in a place where the direction of the hand is always the same with respect to the body, such as placing the left or right palm on the chest. You can use the orientation when you place. Also in this case, the normalization can be performed by the same method.

[Equation 3]

FIG. 9 is a diagram showing expansion / contraction of data by normalization of physique. In the present embodiment, in addition, in order to absorb the difference in the size of the movement caused by the physique and the displacement between the body and each part of the body, the hands when the head, shoulders, navel, and arm are straightened forward. The position of is used to normalize the position of the hand. Therefore, first, X as shown in (Equation 4)
The ratios of the axes, the Y-axis and the Z-axis direction, rx, ry and rz are obtained.

[Equation 4] The ratio obtained by (Equation 4) is, as shown in FIG. 9, the top 911 and the right shoulder 91 of the person who inputs the data of the sign language word dictionary 6.
2, squares 931 and 932 formed by the left shoulder 913, the navel 914, and the hand position 915, and the crown 921, the right shoulder 922, the left shoulder 923, and the navel 924 of the person who has input the sign language for recognition.
It is the ratio of the lengths of the sides of the squares 941 and 942 made at the hand position 925. In this formula (Equation 4), the ratio of the body size of the person who inputs the sign language for recognition to the body size of the person who inputs the data of the sign language word dictionary 6 in the X-axis, Y-axis, and Z-axis directions. Is represented. Since the ratio of the size of the hand position on each coordinate axis of the data of the sign language word dictionary 6 and the newly input sign language data for recognition is the same as the ratio represented by (Formula 4), Data (X3, Y3, Z) that absorbs changes in data due to physique by performing normalization
3) can be obtained.

[Equation 5]

FIG. 10 is a view showing the mounting position of the magnetic field generator in which the normalization of the direction can be omitted. The sign language data obtained by the conversion described in FIGS. 5 to 9 is used for the recognition of the sign language as the normalized data. By performing similar conversion on the sign language data of the sign language word dictionary 6 in advance, it is possible to perform recognition that does not depend on the physique, the position where the sign language is performed, and the direction. Further, in the well-known sign language input device using gloves, the position and direction are detected by installing a device for generating a magnetic field at an arbitrary place and detecting the magnetic field by a sensor attached to the glove. Therefore, a device for generating a magnetic field is installed on the body 101 of a person who performs sign language as shown in FIG. 10, or a chair 10 for sitting for sign language.
When installed in No. 2, the relationship between the sensor and the magnetic field generator is independent of the direction of sign language. That is, normally, the magnetic field generator is placed at a position other than the origin and this is detected by the detector. However, in FIG. 10, since the magnetic field generator is arranged at the origin, coordinate conversion (normalization) with respect to the origin is performed. There is no need to do it. This makes it possible to omit the normalization according to the direction of sign language, that is, the normalization by (Equation 3). Furthermore, if the device that generates the magnetic field is installed at the top of the sign or the navel of a person who performs sign language, such as the reference point that defines the physique, the position can be used as the origin for normalization by the physique. , (Equation 1), there is no need to normalize the position.

The sign language data normalized by (Equation 1) to (Equation 5) is then input to the boundary detection device 3 and the sign language data storage device 4 shown in FIG. The operation of the boundary detection device 3 will be described with reference to FIGS. 11 to 13. Figure 11
FIG. 6 is a diagram showing a relationship between a home position and its detection range. The boundary detection device 3 detects the time at the boundary of the sign language words forming the sign language sentence. When inputting sign language, if it is determined that the position of the hand always returns to a specific position, that is, the home position at the boundary of the sign language words, the boundary detection device 3 always detects the sign language data of the input sign language data at each time. Check the position of. Then, only when the input hand position exists within a certain range from the home position for a certain period of time, it is recognized as the sign language word boundary. Set the home position coordinates to (x0, y0, z
0), the position coordinate of the sign language data at a certain time t is (x
(T), y (t), z (t)), and letting THp be the range in which the home position is determined,

[Equation 6] The case where the following time continues for a certain time Tt is detected as a sign language word boundary. That is, in FIG. 11, the sphere 11 centered on the home position 111 and having a radius THp
The case where 2 includes the position coordinates is detected. Then, the time when the sign language word boundary is detected is sent to the sign language data storage unit 4 as a boundary detection signal d2.

FIG. 12 is a view showing a mounting position when the sign language word boundary detecting switch is mounted on the body.
As a parameter for boundary detection, not only the position of the hand but also the direction of the specific hand or the shape of the specific hand can be used.
The detection method in this case can be performed by detecting the data input by the data grove with the sign language word boundary detection device in the same manner. In addition, when sign language is performed, it can be decided that the hand always stops at the boundary of the sign language words. In this case, the velocity of the sign language data at each time may be calculated, and the case where the velocity is equal to or less than a certain value continues for a certain time may be recognized as a boundary. That is, if the position coordinates of the sign language data at a certain time t are (x (t), y (t), z (t)) and the speed threshold is THv,

[Equation 7] It is only necessary to detect the case where the following time continues for a certain time Tv as a sign language word boundary. In addition, not only the position of the hand, but also the direction and shape of the hand, the speed is calculated in the same way.
You may use this. Further, as the boundary detection device 3, as shown in FIG. 12, the switch 12 attached to the foot is used.
It is also possible to detect the boundaries of the sign language words by using the switches 122 attached to 1 or the neck and inputting these switches as detection signals. The position where the switch is attached may be anywhere as long as it can be moved regardless of the sign language expression. In the left side of FIG. 12, the sign language person detects the boundary of the sign language words by stepping on the switch 121 with his / her foot. Further, on the right side of FIG. 12, the switch 122 is attached to the side of the neck so that the person who is performing the sign language detects the boundary of the sign language words by placing the side of the neck.

FIG. 13 is a diagram showing a structure of a boundary detection device for detecting a sign language word boundary by a facial expression. As shown in FIG. 13, the television camera 13 is used as the boundary detection device 3.
1 and the face image recognition device 132 are used to perform a specific facial expression at the sign language word boundary, and the recognition device 132 recognizes the face image input from the television camera 131 to detect the boundary of the sign language words. You can also As a facial expression recognition method in the face image recognition device 132, an existing technique (for example, the above-mentioned Japanese Patent Application No. 4-247285 and drawings) may be used. When the boundary of the sign language words is detected, the boundary detection device 3 sends the detection signal d2 to the sign language data storage device 4. The sign language data storage device 4 stores the normalized sign language data sent from the normalization device 2. At this time, when the detection signal d2 is sent, the sign language data stored until then is sent to the sign language recognition device 5. As a result, the sign language data storage device 4 stores the sign language data between the detection signals, that is, the data for each sign language word, and sends it to the sign language recognition device 5.
The sign language recognition device 5 uses the sign language data for each sign language word input from the sign language data storage device 4 and the sign language word data stored in the sign language word dictionary 6 to determine what sign language the input sign language is. Recognize As a method for recognizing sign language, there is an existing technique (for example, Japanese Patent Application No. 4-23
No. 5633 specification and drawings) may be used. The result of recognizing the sign language is sent to the output device 7 and is output as characters or voice.

In the above embodiment, only the boundaries between the sign language words forming the sign language sentence are detected, but in addition to this, the start and end points of the sign language sentence can be detected using the same method. it can. That is, whether to perform a motion according to a specific position, direction, or shape at the start point / end point of the sign language sentence, stop the movement of the hand (see FIG. 11), or input with a switch attached to the body (see FIG. 12). Alternatively, a specific facial expression may be performed (see FIG. 13). When a specific position, direction, or shape is used, or when a specific facial expression is used, if a position, direction, shape, or facial expression different from the position, direction, shape, or facial expression for boundary detection of sign language words is used, It is possible to detect a sign language word boundary and a start point / end point of a sign language sentence by using different detection signals corresponding to each. The same input signal is used when detecting the boundaries of sign language words and detecting the start and end points of a sign language sentence at the same position, direction, shape, or facial expression, when stopping motion, or when inputting with a switch. Therefore, the duration of the detection signal may be used. That is, if THw is the threshold value of the duration time for detecting the sign language word boundary, THs is the threshold value of the duration time for detecting the start point / end point of the sign language sentence, and t is the duration time of the detection signal, then (Equation 8) Under such conditions, the boundaries of the sign language words and the start and end points of the sign language sentence can be detected.

[Equation 8] As a result, it is possible to accurately recognize the sign language sentence.

[0019]

As described above, according to the present invention,
Before recognizing, sign language sentences can be divided into sign language words, which can reduce false recognition due to sign language word boundaries.
It is possible to accurately recognize the sign language sentence. Also,
By normalizing the sign language data before inputting the sign language, it is possible to absorb changes in the sign language data due to the physique of the person who performs the sign language and the posture in which the sign language is used. It becomes possible to make a good recognition of.

[Brief description of drawings]

FIG. 1 is a diagram showing an overall configuration of a sign language interpreting apparatus showing an embodiment of the present invention.

FIG. 2 is a diagram showing a structure of sign language data in the present invention.

FIG. 3 is a diagram showing a structure of data stored in a sign language word dictionary in FIG.

FIG. 4 is a diagram showing reference points input for normalizing a position and a physique.

FIG. 5 is a diagram showing conversion of a coordinate system by position normalization.

FIG. 6 is a diagram showing a vector representing a direction of a hand.

FIG. 7 is a diagram showing the relationship between the direction of the hand and the coordinate system when the hand is placed in front of the body.

FIG. 8 is a diagram showing a relationship between a body direction and a coordinate system.

FIG. 9 is a diagram showing expansion / contraction of data by normalization of physique.

FIG. 10 is a diagram showing a mounting position of a magnetic field generator in which direction normalization can be omitted.

FIG. 11 is a diagram showing a relationship between a home position and its detection range.

FIG. 12 is a diagram showing an example of attachment when a switch for sign language word boundary detection is attached to the body.

FIG. 13 is a diagram showing a configuration of a boundary detection device when detecting a sign language word boundary by a facial expression.

[Explanation of symbols]

1 Sign Language Input Device, 2 Normalization Device, 3 Boundary Detection Device, 4 Sign Language Data Storage Device, 5 Sign Language Recognition Device, 6
Sign language dictionary, 7 output device, 41 crown, 42 right shoulder, 43 left shoulder, 44 navel, 45 hand position when extended, 61 vector perpendicular to palm, 62 palm fingertip direction vector, 63 palm Lateral direction vector,
71 palm direction, 72 palm lateral direction, 101 body position, 102 chair, 111 home position, 11
2 balls, 121, 122 switch, 131 TV camera, 132 Face image recognition device.

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yu Oki 1-280, Higashi Koikeku, Kokubunji, Tokyo, Central Research Laboratory, Hitachi, Ltd. (72) Eiji Ohira 1-280, Higashi Koikeku, Kokubunji, Tokyo Hitachi, Ltd. Central Research Center

Claims (16)

[Claims] 1. A sign language input means for inputting sign language data obtained by converting the position of a hand in sign language into an electric signal, a sign language recognition means for recognizing a sign language word from the input sign language data, and a recognition result of the sign language word. A sign language interpreting device, comprising: an output unit for outputting the boundary and a boundary detecting unit for detecting a boundary between the sign language words forming the sign language sentence. 2. The sign language interpreting device according to claim 1,
The boundary detection means sets a predetermined position, direction, shape, or gesture of the hand, or a combination thereof, and the hand that is always set at the boundary of all the sign language words forming the sign language sentence. A sign language interpreter characterized by detecting the boundaries of sign language words by performing position, direction, shape, or gesture. 3. The sign language interpreter according to claim 1,
The sign language interpreting device, wherein the boundary detection means detects the boundary of the sign language word by detecting the boundary of the sign language word without stopping the motion of the hand and detecting the stop of the motion of the hand for a predetermined time. 4. The sign language interpreter according to claim 1,
The boundary detecting means has a switch installed on the floor of the body of the person who performs sign language or on the floor near the person who performs sign language, and when the person who performs the sign language determines that it is the boundary of the sign language words, the switch that is installed is used. A sign language interpreter characterized by detecting the boundaries of a sign language word by sending a signal. 5. The sign language interpreter according to claim 1,
The sign language interpreting device, wherein the boundary detecting means performs a predetermined facial expression when the person who performs the sign language determines the boundary of the sign language word, and detects the boundary of the sign language word by detecting the facial expression. 6. A sign language input means for inputting sign language data obtained by converting the position of a hand of sign language into an electric signal, a sign language recognition means for recognizing a sign language word from the input sign language data, and a recognition result of the sign language word. And a sign language sentence boundary detection means for detecting the start point and the end point of the sign language sentence. 7. The sign language interpreting device according to claim 6,
The sign language sentence boundary detection means sets a predetermined hand position, direction, shape, or gesture, or a combination thereof, and the hand position and direction that are always set at the start point and the end point of the sign language sentence, A sign language interpreting device, which detects a start point and an end point of a sign language sentence by performing a shape or a gesture. 8. The sign language interpreter according to claim 6,
The sign language sentence boundary detection means detects the start point and the end point of the sign language sentence by always stopping the hand movement at the start point and the end point of the sign language sentence and detecting the stop of the hand movement for a predetermined time. A featured sign language interpreter. 9. The sign language interpreter according to claim 6,
The sign language sentence boundary detection means has a switch installed on the body of the person who performs the sign language or on the floor near the person who performs the sign language, and is installed when the person who performs the sign language determines the start point and the end point of the sign language sentence. A sign language interpreting device, which detects a start point and an end point of a sign language sentence by sending a signal using a switch. 10. The sign language interpreting apparatus according to claim 6, wherein the sign language sentence boundary detecting means performs a predetermined facial expression at the time when the person who performs the sign language determines the start point and the end point of the sign language sentence, and displays the facial expression. A sign language interpreter characterized by detecting the start and end points of a sign language sentence. 11. A sign language input means for inputting sign language data obtained by converting the position of a hand in sign language into an electric signal, and normalizing the input sign language data in order to absorb changes in the data due to the physique of the person who performs the sign language. A sign language interpreting device comprising: a normalization means, a sign language recognition means for recognizing a sign language word from input sign language data, and an output means for outputting a recognition result of the sign language word. 12. The sign language interpreting apparatus according to claim 11, wherein the normalizing means inputs a small number of positions as a reference in advance, so that the sign language depends on the body size of the person performing the sign language or the position where the sign language is performed. A sign language interpreter characterized by normalizing data differences. 13. The sign language interpreting apparatus according to claim 12, wherein the normalizing means includes a head position, left and right shoulder positions, waist positions, navel positions, and hands when the arm is straightened forward. A sign language interpreting device characterized by normalizing the difference in the sign language data depending on the size of the body of the person performing the sign language or the position where the sign language is performed by inputting the position. 14. The sign language interpreting apparatus according to claim 11, wherein the normalizing means normalizes a difference between the sign language data depending on the direction in which the sign language is performed by inputting a reference hand direction in advance. A featured sign language interpreter. 15. The sign language interpreting apparatus according to claim 14, wherein the normalizing means normalizes the difference in the sign language data depending on the direction in which the sign language is performed, from the direction of the hands when both hands are brought together in front of the body. A sign language interpreter characterized by the above. 16. The sign language interpreting apparatus according to claim 11, wherein the normalizing means installs a means for detecting a position and a direction on a body of a person who performs sign language or a chair which is sitting for performing sign language. A sign language interpreter characterized by normalizing the difference in sign language data depending on the position and direction of sign language.