JPH07105848B2 - Media conversion method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Field of Industrial Application) The present invention relates to a media conversion system in a workstation for multimedia.

(Prior Art) With the development of information processing technology, various workstations have been developed. A workstation has been developed which can process not only general code information but information of various media such as voice data and image data. The communication media are also diversifying.

By the way, this kind of multimedia-compatible workstation is installed in each office, and is used only in that office. In recent years, a plurality of workstations are connected to each other via a communication network, information is communicated between the plurality of workstations, and the processing functions of the workstations are used from various information terminals.

However, when communicating information between workstations of this type, or between workstations and other information terminals,
It is premised that the media that they can handle are the same, that is, the functions of the terminals that perform information communication are the same.
However, it is almost impossible to provide all the information terminals that have the possibility of communicating information with the workstation with the same processing function as the workstation. Therefore,
The communication partner terminals with which the workstation can communicate information are limited, and the expansion of the communication range is greatly restricted.

(Problems to be Solved by the Invention) The present invention has been made in view of the problem that the function of the workstation cannot be effectively used from the information terminal unless the information terminal has the same function as the workstation. It was done and the purpose is to
It makes it possible to convert the information to be communicated from the workstation to the communication partner terminal into a medium according to the type of media that can be processed by the communication partner terminal and to perform information communication, and to change the information held by the workstation into various information. It is to provide a media conversion method that can be output on a terminal.

[Structure of the Invention] (Means for Solving Problems) The present invention, as schematically shown in FIG. 1, includes a voice recognition device, a character recognition device, a figure recognition device, an image recognition device,
A voice synthesizing device, an image synthesizing device, a graphic synthesizing device, a plurality of types of media converting units A such as various data compressing / expanding devices, and a communication device C that performs information communication with various communication partner terminals B. In the equipped multimedia-compatible workstation, the input medium of information to be used for communication is detected by the input medium detection unit (own-station media identification unit D), and the partner media detection unit (remote-station media identification unit) is detected. In E), a communication medium for processing the above information is detected by detecting a medium that can be processed by the communication partner terminal, and a media conversion table G is obtained according to the input medium and the communication medium under the control of the medium selection control unit F. The type of media conversion required for communicating the input information to the communication partner terminal is obtained by referring to the media conversion type according to the type of media conversion. The input information is converted into media and transmitted to the communication partner terminal.

Specifically, for example, when the communication partner terminal is a telephone terminal equipped with a facsimile device, the type of the communication partner terminal is detected, it is requested that image data and voice data can be communicated, and the media of this communication data is used. Accordingly, for example, the document information represented by the character code string is media-converted into character image information for information communication.

Further, the media selection control section F refers to the self media function table H as needed to obtain the media conversion function and the like provided in the workstation, and controls the selection of the media conversion type.

(Operation) According to the present invention, even if the communication partner terminal does not have the same processing function as the workstation, the partner media detection unit of the workstation determines the type of media that the communication partner terminal can process, for example, the communication partner terminal. From the media identification signal sent from the communication device, and according to the detection result, the media conversion means is selectively activated to obtain the type of media suitable for information communication with the communication partner terminal, and the media of the input information is communicated. Communication is performed by converting to media that the other terminal can process. Therefore, it becomes possible to convert the information medium possessed by the workstation according to the processing function (processing medium) of the communication partner terminal, and effectively output this by communication.

For example, when the communication partner terminal connected to the workstation is a facsimile machine, it becomes possible to communicate and output the document information represented by the character code string as a character image, and when the telephone terminal is connected. It becomes possible to communicate and output the document information as a synthetic voice.

Further, in the workstation, for example, it becomes possible to recognize the information input as a character image into a character code string, convert the information of the character code string into voice information, and output the information to the telephone terminal by communication.

Embodiments Embodiments of the present invention will be described below with reference to the drawings.

FIG. 2 is a schematic configuration diagram of an intelligent workstation according to the embodiment of the present invention. This intelligent workstation
It is configured to include the following parts.

Bus 1; used for performing necessary information transfer between the units described below.

Control unit 2; which is mainly composed of a microprocessor and controls the operation of each unit of the intelligent workstation.

Image input device 3; consisting of a camera, scanner, OCR, etc., for inputting various image information.

Position coordinate input device 4; comprising a tablet, mouse, etc., for inputting specified position coordinate information.

A voice input unit 5 is composed of a microphone or the like and inputs voice information.

Keyboard section 6; provided with a plurality of keys for inputting character / symbol codes, control codes, and the like.

IC card section 7; an IC card is mounted as described later
The necessary information is input to and output from the card.

Bus controller 8: Controls information transfer between each unit via the bus 1.

Audio output unit 9: A speaker or the like, which outputs audio information.

Display unit 10: A CRT display, a liquid crystal display, or the like, which displays characters, figures, images, and the like.

Image output device 11; consisting of fax, color printer, etc.
Print out various image information.

Communication devices 12, 13: Performs information communication between the work station and a telephone, or another work station or terminal installed in a remote place.

Switching device 14; switches and uses a plurality of communication devices.

Timer unit 15; Provides time information and time information to the workstation.

Encryption processing unit 16; Performs encryption processing on various information.

Voice collating unit 17; collates whether or not the given voice information is a specific voice.

Image collating unit 18; collates whether or not the given image information is a specific image.

Voice recognition unit 19; recognizes the given voice information.

Voice analysis unit 20; analyzes the voice by extracting features of the voice input from the voice input unit 5 or the like.

Character recognition unit 21; recognizes character / symbol patterns input from the image input device 3 or the like.

Image recognition unit 23; recognizes a graphic image or the like input from the image input device 3 or the like.

Output form selection unit 24: Selects and controls the form of information output from the workstation.

Work environment data collection unit 25: Collects and inputs information such as the functional status of the workstation and the work environment in the office due to the function status.

Speech synthesizer 26; Generates synthetic speech according to the processed data.

Image synthesizing unit 27: Performs a synthesizing process of a plurality of image information and executes an image editing process according to the process data.

Figure synthesizing processing unit 28: synthesizes various figures, and executes edit processing such as addition / deletion of figures according to processing data.

Audio compression / decompression unit 29: Compresses and encodes audio data, and restores and expands compressed audio data.

Image compression / decompression unit 30; compresses and encodes image data, and restores / decompresses compressed image data.

The signal processing unit 31 executes a series of signal processing such as coding / compression of various signal information, decompression / decompression thereof, addition of necessary information, and the like.

Database unit 32: Classifies various information into a plurality of relations and stores them as a database. It should be noted that this database is constructed not only as code information but also as an image or a voice machine.

The intelligent workstation according to the present invention is basically configured to include the above-described units, and effectively utilizes the functions of the above-described units to exhibit intelligent functions as a whole. .

Next, the IC card unit 7, the encryption processing unit 16, and the like, which are not general like the keyboard unit 5 and the like described above, and have a characteristic function in this intelligent workstation will be described in more detail.

First, as shown in FIG. 3, for example, an IC card has a built-in semiconductor circuit such as a microprocessor and a memory circuit in a card body 7a having a size of a business card. An interface section 7b for connecting to the station body and a display window section 7c are provided.

The display window portion 7c is formed by embedding a transparent polarizer, and its position is set so that it does not overlap with the interface portion 7b or the semiconductor circuit. Further, in the card body 7a, only the portion corresponding to the display window portion 7c may be transparent, or the entire substrate thereof may be transparent.

Therefore, as shown in its exploded perspective view in FIG. 4, the IC card has a pair of cover substrates 7d and 7e, an embedded substrate 7f sandwiched between these cover substrates 7d and 7e, and a core sheet material 7g. The printed circuit board 7h is integrally thermocompression bonded.

An input / output terminal 7i is provided at a position facing the interface portion 7b of the printed board 7h, and a display window portion 7c is provided.
A liquid crystal display device 7j is provided at a position opposed to. Furthermore, a semiconductor integrated circuit 7k can be provided on the printed board 7h.
Further, the cover substrate 7c is provided with a metal foil 7m for dissipating the heat generated in the printed circuit board 7h.

The cover substrates 7d and 7e, the embedded substrate 7f, the core sheet material 7
The holes formed in g are provided at positions facing the semiconductor integrated circuits 7j and the like integrated on the printed board 7h, respectively. The semiconductor integrated circuit 7k is placed in these holes.
The cover substrates 7d and 7e, the embedded substrate 7
f, the core sheet material 7g, and the printed board 7h are laminated and integrated to form an IC card. The input / output terminal 7i is exposed through a hole formed in the cover substrate 7d and is electrically connected to the workstation main body.
Make up.

The liquid crystal display device 7j has a liquid crystal layer sandwiched between a pair of polyether sulfone film substrates provided with spacers, as shown in the sectional structure of the printed circuit board 7k portion in FIG. 5, for example. A transparent conductive film is formed on the inner surface of the film substrate, and a polarizer or reflector is provided on the film substrate on the lower surface side. In this way, using a polyether sulfone film substrate, a liquid crystal display device 7j
With this configuration, it is easy to reduce the thickness to 0.6 μm or less, and the IC card itself can be made thinner than when a liquid crystal display device is configured using a glass substrate.

Further, the driving power of this IC card may be supplied from the workstation main body side via the interface section 7b, or may be built in the card. In this case, for example, a sheet-shaped battery using a polymer film may be incorporated.

Therefore, the semiconductor integrated circuit 7k includes, for example, as shown in FIG. 6, a CPU 7p, a data memory PROM 7q, an E ² PROM 7r,
And a selection unit 7s for these memories. The PROM 7q is a large-capacity non-volatile memory that cannot be erased / rewritten, and stores a control program for the CPU 7p, information to be permanently recorded, and the like. Also E ² PROM
Reference numeral 7r is a rewritable small-capacity non-volatile memory, and stores information such as a transaction number of information and information updated when the information is used.

These memories are selectively driven by the control of the selecting unit 7s, and input / output of information with the CPU 7p. CPU
7p uses these memories to perform the necessary information processing,
Further, information is input / output from / to the intelligent workstation main body through the interface section through the terminal section 7i described above.

The IC card unit 7 is equipped with such an IC card,
Information will be input to and output from the IC card.

It goes without saying that the IC card is not limited to the above-mentioned configuration, and the IC card unit 7 may be used depending on the configuration.
Needless to say, is configured.

Next, the encryption processing unit 16 will be described.

The encryption processing unit 16 is, for example, as shown in FIG.
16a, a decryption unit 16b, a private key file unit 16c, a public key file unit 16d, and a key update unit 16e.

Then, as shown in the concept in FIG. 8, the given communication original text is encrypted according to the encryption key to generate the encrypted communication text, and conversely, the given encrypted communication text is decrypted according to the encryption key and Executes the process for obtaining the original text.

The private key file unit 16c and the public key file unit 16d store the keys used for this encryption / decryption, and the key updating unit 16e controls the updating of these filed keys.

Here, the secret key is a key known only to the workstation that owns the encryption processing unit 16, and is kept secret from other workstations. On the other hand, the public key forms a pair with each secret key set in each workstation, and is given to other workstations and made public. The public key file unit 16d stores the public keys published by each of these workstations in association with each workstation.

As shown in FIG. 9, the encryption unit 16a includes an RSA processing unit 16i and an encryption type addition unit 16j. Then, when attempting to communicate information by encrypting the communication source text, the communication source text is encrypted using the public key disclosed by the workstation of the communication partner, and information indicating the type of encryption is added to the encrypted communication text. Communication information is created and communicated. The encryption type information is, for example, "0".
It is composed of information indicating that it is not encrypted in step 1, and that it is encrypted in step 1, and information indicating the encryption method.

Further, the decryption unit 16b inputs the encrypted communication text which a certain workstation using the public key published by the own workstation encrypts and communicates, and decrypts this using a secret key corresponding to the secret key. As shown in FIG. 10, it comprises a ciphertext division unit 16k, an encryption type determination unit 16m, switching units 16n and 16p, and an RSA processing unit 16q.

The ciphertext division unit 16k is for dividing the communication information transmitted in the above-mentioned format into the above-mentioned encryption type information and encrypted communication text, and the encryption type determination unit 16m uses the encryption type information to communicate the information. It is determined whether or not the sentence is encrypted. If it is not encrypted, the communication text is output via the switching units 16n and 16p, and if it is encrypted, the communication text is guided to the RSA processing unit 16q. The RSA processing unit 16q decrypts the encrypted communication text using the secret key, and outputs the decrypted communication text via the switching unit 16p.

The RSA processing units 16i and 16q are configured to include, for example, a block dividing unit 16s, a power / residue calculating unit 16t, and a block connecting unit 16u, as shown in FIG.

Here, the block division unit 16s divides the given signal sequence into blocks Mi having a constant length, and the power / residue calculation unit 16t uses Ni = Mik for each block Mi using the encryption key k. (Mod n) signal sequence Ni is obtained. However, n is a fixed value. This signal series Ni is sequentially connected and output via the block connecting section 16u.

In the encryption processing, the above-mentioned signal sequence Mi is the original communication text, and the encrypted communication text from this original communication text is the signal series Ni.
Is required as. Further, in the decryption process, the signal series Mi is an encrypted communication text, and the communication original text decrypted from this encrypted communication text is obtained as the signal series Ni.

The key k responsible for such encryption / decryption is the above-mentioned public key and secret key, which are set in pairs.

Therefore, a workstation can encrypt communication information according to a public key published by another workstation, but can decrypt the encrypted communication text only by a secret paired with the public key. Only certain workstations can know the key.

Therefore, a workstation that attempts to communicate by encrypting certain information encrypts the original communication text according to the public key disclosed by the workstation of the communication partner and communicates.
Then, the communication information can be decrypted only by the communication partner workstation having the secret key.

Note that it is not necessary to store all the public keys published by other workstations in the public key file 16d. For example, in a public key file memory provided separately for the system, the public key published by each workstation is filed in association with each workstation. When information communication becomes necessary, the public key of the communication partner may be read from the public key file memory and stored in the public key file unit 16 of the own workstation.

The above is the basic configuration and function of the encryption processing unit 16.

Next, the image matching unit 18 will be described.

The image matching unit 18 inputs the image information input from the image input device 3, for example, the image of the face of an individual, and identifies the individual.

Figure 12 shows the schematic structure of this image matching unit.
Reference numeral 18a is an image storage unit, 18b is a normalization circuit, 18c is a binarization (thinning) circuit, and 18d is a characteristic data extraction circuit. Further, 18e is a data storage unit that stores the image data, 18f is a search circuit, 18g is a matching circuit, and 18h is an output unit.

The image storage section 18a stores the image information input through the image input device 3 and uses it for the image matching process. The normalization circuit 18b normalizes the image information stored in the image storage unit 18a, and the binarization circuit 18c binarizes the image information. Specifically, here, the normalization circuit 18b normalizes the size of the face in order to identify the individual from the image of the face. The binarization circuit 18c performs edge line segment detection, thinning processing of the edge line segment, etc. on the normalized face image to obtain a binary image of the image.

The feature data extraction circuit 18d thus normalizes.
The feature data is extracted from the binarized image information. That is, in the collation processing using the image of the face, for example, as shown in FIG. 13, the contour of the face is extracted as one feature, and the features such as eyes, nose, and mouth in the image are extracted. . Specifically, the contour features of the face are extracted as classified code information, and the distance 1 between the eyes, the size m of the mouth, the distance n between the eyes and the like are extracted as numerical data features of the image. is doing.

In the data storage unit 18e, the feature data of the image of the face obtained for each individual is registered in advance as shown in FIG. 14, for example. That is, the feature data of the face image described above is registered for each individual with the individual name as an identification name, and the image data of the face is linked by a pointer.

The search circuit 18f searches the data storage unit 18e based on the feature data extracted by the feature data extraction circuit 18d. Then, the search data is given to the matching circuit 18g,
It is collated with the characteristic data obtained by the characteristic data extraction circuit 18d.

For example, when the characteristic data of the input image obtained by the characteristic data extraction circuit 18d is Xi (i is a characteristic type) and the characteristic data of the image registered in the data storage unit 18e is Yi, Is performed, and the one with the smallest value of the calculation result D is identified as this individual. This identification result is output via the output unit 18h.

The image collating unit 18 basically collates the input image in this way, and performs, for example, individual identification of the input image.

Next, the voice recognition unit 19 will be described.

The voice recognition unit 19 is configured, for example, as shown in FIG. The voice input circuit 19a inputs a voice signal input from the voice input unit 5 or a voice signal received by the communication devices 12 and 13 via a public telephone line. It is composed of an amplifier that amplifies to various signal levels, a bandpass filter for band limitation, an A / D converter, and the like. The input voice is this voice input circuit 19
At a, for example, the signal is limited to a signal in the frequency band of 30 to 3400 Hz, and is quantized into a 12-bit digital signal at a sampling period of 12 KHz.

The sound processing unit 19b is composed of, for example, a sum-of-products circuit configured by dedicated hardware. Then, basically, it operates at high speed in a pipeline manner in synchronization with the voice input circuit 19a.

The acoustic processing here is executed by two types of band pass filter groups. One of them is a 16-channel filter bank, through which changes in the spectrum of the input audio signal are extracted. Now one has the same band, 4
This is a gross filter divided into channels, and the acoustic features of the input voice are extracted through this gross filter.

These two types of filter groups (filter bank and gross filter) are configured as, for example, a fourth-order cyclic digital filter. Then, for example, the filtering output is obtained every 10 msec. The control of the acoustic processing unit is performed by a micro program method.

Thus, the preprocessing / recognition unit 19c includes a high speed processor 19d, a pattern matching processing unit 19e, a word dictionary memory 19f, and a buffer memory 19g.

The buffer memory 19g receives the audio signal filtered by the acoustic processing unit 19b and stores, for example, a maximum of 1.8 seconds of audio data. The high speed processor 19d detects the voice section, resampling, labeling, with respect to the data stored in the buffer memory 19g.
The recognition process by the transition network and the comprehensive logic judgment process are executed. Further, the high speed processor 19d performs communication with the host computer and operation control of the entire voice recognition unit 19.

For the voice data processed by the high speed processor 19d, the pattern matching processing unit 19c uses the word dictionary memory 1
Matching processing such as complex similarity calculation is executed with the standard pattern data of the word voice registered in 9f to obtain the recognition candidate.

For example, speech words to be recognized are uttered discretely.
Therefore, the high-speed processor 19d uses, for example, 10 m for sound processing.
The input section of word speech is detected using the input speech energy calculated every sec.

Specifically, as shown in FIG. 16, a threshold E _θ adaptively calculated from the background noise level and the input voice level is used,
When the input voice signal level exceeds the threshold value E _θ for a certain period of time or longer, the time when the input voice signal level exceeds the threshold value E _θ is detected as the beginning S of the voice word. After that, when the level of the input voice signal is continuously lower than the threshold E _θ for a certain period of time or more, the time point when the level is lower than the threshold E _θ is the end E of the voice word
Is detected as.

By the way, speech recognition can be considered as a kind of pattern recognition. However, pattern variations peculiar to voice, individual differences caused by the gender of the speaker, the shape of the vocal organs, vocalization method, etc., the noise generated by the speaker himself or the surrounding environment, and in the case of telephone voice, the public There are problems of level difference and noise due to passing through the telephone line. Therefore, in consideration of these, there is a problem of how to accurately and stably recognize the voice by absorbing the above-mentioned variable elements.

Therefore, in this preprocessing / recognition unit 19c, the pattern matching method and the structure analysis method are combined in two stages, and a recognition method called a hybrid structure matching method is adopted.

That is, when the word voice section is detected as described above, first, the voice section (S, E) is divided into 15 equal parts, and the 16 points are respectively set as resample points. Then, the spectrum at each resample point is extracted from the 16-channel audio data (spectrum time series) that has been acoustically processed as described above. still,
If there is a deviation between the sample point of the audio data and the resample point, the spectrum of the nearest point of the resample point may be extracted.

By this re-sampling processing, a 16 × 16 (= 256) -dimensional voice pattern vector X is obtained. That is, j-th (j = 1,
(2,3, ~ 16) Let rj be the resample point,
16 channels spectral data _{S (rj) = (S 1} rj, S 2 rj, ~S 16 rj) respectively obtained as, X = and reordering these _{Sirj (S 1 r 1, S} 1 r 2, ~S A vector X of a voice pattern of ₂ r ₁ to S ₁₆ r ₁₆ ) t is obtained. However, t indicates the transpose of the matrix.

The input voice pattern vector X thus obtained
And the similarity with the standard pattern of the word voice registered in advance in the word dictionary memory 19f is calculated by, for example, the composite similarity method.

Standard patterns where pre-registered word speech the word dictionary memory 19f, the the word category _{ωk (ψ 1 k, ψ 2} k, ~ψ L k) (λ 1 k, λ 2 k, ~λ L k However, it is prepared as (λ ₁ k ≧ λ ₂ k ≧ ˜ ≧ λ _L k). Note that ψ, k, λ, k is the category ωk
Are the eigenvectors and their eigenvalues in the distribution matrix K of the pattern vector X belonging to For such a word dictionary, the above-mentioned composite similarity S (k) is Calculated as In the above equation, ‖X‖ is the norm of the vector X.

Such a composite similarity calculation is performed for each category, and the similarity value positioned at the top and the category name that obtained it are obtained as a pair.

By the pattern matching by the composite similarity method as described above, it is possible to perform the recognition processing that rescues many pattern variations. However, in a similar pattern or a pattern in which noise is added, the difference in the similarity value between different categories may be small.

Therefore, as a complement to the pattern matching method as described above, the following structural analysis method is introduced. This structural analysis focuses on the differences in the sounds that make up the word speech, and performs recognition processing. It consists of a phoneme label series and an acoustic feature series.
It uses two time series.

That is, the phoneme label sequence is obtained by calculating the similarity with the phoneme dictionary using the 16-channel spectrum calculated every 10 msec from the input speech signal, and labeling the phonemes having a certain value or more. The phoneme label is composed of, for example, 6 types of 5 vowels and nasal sounds. At this time, it is desirable to prepare the phoneme dictionary separately for male voice and female voice.

Here, it is more difficult to individually label consonants as phonemes, as compared to vowels that are sounded relatively stably. So for that consonant we label its acoustic features,
This is the characteristic information. Specifically, the acoustic feature is extracted from the output of the 4-channel gloss filter and the sound energy obtained by the acoustic processing. The acoustic features that are feature-extracted and labeled in this way are, for example, as shown in FIG. 17 in comparison with the features of the output of the gross filter,
It consists of 12 types such as silence, voicelessness, friction, bursting, and energy dip.

Then, the phoneme / acoustic label sequence obtained for the input speech is generated for each word category over the range including the speech section (S, E), and is converted into a transition network as shown in FIG. 18, for example. Is entered.

The presence or absence of a designated phoneme label or acoustic feature is checked for each node of this transition network. If it is not present, reject it, and if it is present, transition to the next node,
The input sequence that reaches the goal of the transition network at the end of the feature sequence is accepted and its category is obtained. It should be noted that the direction of checking the sequence can be selected from forward and backward for each network.

The comprehensive judgment logic is a logic for making a final judgment by combining the candidate categories ordered by the pattern matching as described above and the recognition result obtained by the transition network.

That is, this comprehensive judgment logic compares this with a predetermined threshold value θ when the maximum similarity S ₁ obtained by pattern matching is set. Then, in the case of (S ₁ <θ), this is rejected as noise.

In the case of (S ₁ ≧ θ), using a different threshold value [Delta] [theta] (S ₁
A category having a degree of similarity equal to or greater than −Δθ) is extracted as a candidate. When the number n of the extracted categories is 1, this is extracted as the recognition result. When a plurality of categories are extracted, the analysis result by the transition network is referred to and only the categories accepted by the transition network are extracted. Then, the category having the highest similarity among them is obtained as the recognition result.

In addition, when the category accepted by the transition network is not included in the categories extracted by the threshold processing,
The judgment cannot be made.

As described above, the input word speech is recognized by integrating the pattern recognition processing result by the composite similarity method and the recognition result using the transition network.

FIG. 19 shows the flow of the word speech recognition processing procedure in this speech recognition unit. After speech section detection processing, resample processing is performed for pattern matching, and at the same time labeling processing is performed for checking by the transition network. Then, it is shown that these recognition results are integrated to perform the comprehensive decision logic process. Such processing is executed under the processing sequence of the high speed processor 19d.

By the way, in the case of recognizing words in continuously uttered voices instead of discretely uttered word voices, the following may be done. That is, in this case, the input voice may be divided into various partial sections, and the word similarity may be obtained by performing word identification for each partial section.

Specifically, for example, as shown in FIG. 20, all the analysis frames in the input speech section are set as boundary candidates of partial sections, and the input speech section is divided into a plurality of partial sections.
At this time, since the maximum time length Dmax and the minimum time length Dmin can be set for the duration time of the word to be recognized, only the partial section within the range may be the recognition processing target.

Here, in the example shown in FIG. 20, two partial sections are obtained on the assumption that the number of words of continuously uttered speech is two. However, since the number of words in the input voice is generally unknown, it is sufficient to detect each of the partial sections on the assumption that 2 to n words exist as word candidates. Then, the word similarity is calculated for each detected subsection,
The connection relationship of the similarity results may be compared with each other to find the boundary of the most reliable subsection, and the word recognition results of the subsections delimited by the boundary may be found.

However, when the word similarity calculation is performed by obtaining the subsections in this way, the number of subsections becomes enormous, which hinders speeding up of processing. Therefore, in consideration of speeding up the processing, for example, the number of input words is 2 to 5 words,
The duration of one word is 128 to 640 msec, the ratio of word length in one utterance is 2.5 or less, and the frame period is 16 msec (8 msec.
A partial interval may be detected by adding a restriction such as (taking out one word every two words in a cycle).

This makes it possible to effectively recognize each word in the continuously uttered voice.

By the way, learning of a dictionary (word dictionary) used for such a voice recognition process is performed as follows.

This learning process is roughly classified into a process for obtaining the characteristic nucleus from the vowel pattern and the consonant pattern, and a process for obtaining the eigenvalue and eigenvector for the characteristic nucleus. Then, N eigenvalues and eigenvectors are obtained in order from the largest eigenvalue. This process is generally called KL expansion.

First, the process of obtaining the characteristic kernel will be described. The characteristic kernel K of the input voice pattern (learning pattern) is obtained as follows when the vertical vector of the learning pattern is Sm.

Here, Sm = (Sm ₁ , Sm ₂ , Smn) t Note that this learning pattern Sm is 64 when it is a consonant pattern.
Given as a vertical vector of dimensions. In the case of a vowel pattern, it is given as a 16-dimensional vertical vector.

Then, the characteristic kernel K is obtained by multiplying, for each of m learning patterns, the vertical vector Sm by the horizontal vector Sm obtained by transposing the vertical vector Sm, and calculating each element of the matrix by the above-mentioned m learning patterns. It is calculated by averaging over. Therefore, the number of elements of the characteristic kernel is the square of the number of elements of the above vector.

Incidentally, in order to obtain the characteristic kernel K reflecting the pattern distribution of the category by such processing, a certain amount of learning pattern is required. Therefore, it is necessary to store a predetermined number of learning patterns in the learning pattern memory in advance.

However, in the case of vowels, it is sufficient to prepare learning patterns of at least 6 categories in 16 dimensions, but in the case of consonants,
There are 101 categories, and it is necessary to obtain them as 64-dimensional data. For this reason, it cannot be denied that a huge memory capacity is required as it is.

Therefore, in order to obtain the characteristic kernel K that reflects the pattern distribution with a small number of learning patterns, the following characteristic kernel update processing is performed, and the characteristic kernel is gradually improved to a form that reflects the pattern distribution by sequential calculation. To

That is, the arithmetic processing of K = K '+ wSnSnt is repeatedly executed. However, w is a weighting factor when updating the characteristic kernel. The weighting coefficient w takes a positive or negative value, and if it is positive, it has the effect of increasing the similarity to the input pattern of the characteristic kernel matrix, and conversely, if it is negative, it has the effect of decreasing the similarity.

K'denotes the characteristic nucleus before learning the learning pattern Sn, and K denotes the characteristic nucleus updated by learning the learning pattern Sn.

After that, for the characteristic nucleus obtained in this way,
A process of obtaining the eigenvalue and the eigenvector is performed, and a standard pattern used for the above-described composite similarity calculation is created based on the eigenvalue and the eigenvector.

The standard pattern is obtained by KL expanding the characteristic kernel, and the standard pattern is obtained by, for example, KL expansion based on the power method.

Now, it is assumed that the characteristic kernel K has eigenvalues λ ₁ , λ ₂ , ... λn and corresponding eigenvectors ξ ₁ , ξ ₂ , ... ξn.
In this case, the arbitrary vector u ₀ is the eigenvector ξ
_By linearly combining ₁ , ξ ₂ , 〜ξn Is represented as At this time, since the relationship of Kξi = λiξi holds, Becomes

here Therefore, if S becomes sufficiently large, the second term of the above equation will converge to zero.

Therefore, the above equation can be regarded as Ksu ₀ = α ₁ λ ₁ sξ ₁ .

This means that the ratio of (Ks ⁺¹ u ₀ ) and (Ksu ₀ ) is eigenvalue λ _1.
Is shown. It is also shown that (Ksu ₀ ) is proportional to the eigenvector ξ ₁ .

By the way, in the calculation process based on such a theory, the intermediate result of the calculation often scales out immediately. Therefore, let u ₀ be an arbitrary unit vector, for example, and execute an operation of vs ₊₁ = Kus us ₊₁ = (vs ₊₁ ) / (bs ₊₁ ) (s = 0,1,2, ...) To do. Here, (bs ₊₁ ) is the element having the largest absolute value of the vector (vs ₊₁ ). At this time, Therefore, λ ₁ , bs ₊₁ , ξ ₁ , us ₊₁ can be obtained from this.

When the eigenvalue λ ₁ having the largest absolute value and the eigenvector ξ ₁ are obtained in this way, the eigenvalue λ ₂ and the eigenvector ξ ₂ having the next largest absolute value are obtained in the same manner.

Here Given _{K '= K-λ 1 ξ} 1 ξ 1 t, ξ 1 tξi = 0 (i = 2,3, ~, n) _{from, K'ξ 1 = Kξ 1 -λ 1} ξ 1 ξ 1 tξ ₁ = λ ₁ ξ ₁ −λ ₁ ξ ₁ = 0 (i = 1) K′ξ ₁ = K ξi −λ ₁ ξi ξit ξi = λi ξi (i ≠ 1). Therefore, it can be seen that K'has an eigenvalue of | λ ₂ |>...> | λr |>...> | λn |> 0. In addition, here, it is assumed that ξi is normalized.

Such processing can be achieved by repeatedly executing the above-described processing on K ′ obtained by converting the characteristic kernel as K ′ = K−λ ₁ ξ · ξt. By this processing, the eigenvalue having a large absolute value and the eigenvector corresponding thereto are sequentially obtained, and the dictionary is learned.

FIG. 21 shows a procedure of dictionary learning processing executed based on such a calculation algorithm.

Next, the character recognition unit 21 will be described.

The character recognition unit 21 includes a first character recognition block for recognizing a character read by a scanner and a second character recognition block for recognizing character information input online via a tablet or the like.
And a character recognition block of.

This first character recognition block is, for example, as shown in FIG. 22, an image memory 21a for storing image data read and input by a scanner and an object to be recognized from the image data stored in the image memory 21a. An area detection unit 21b that detects an area in which characters are written, and a character extraction unit 21 that extracts character data to be recognized from the image data stored in the image memory 21a according to the area detection result.
c, and a recognition unit 21e for recognizing characters by individually collating the standard character patterns of the recognition target characters registered in advance in the standard pattern dictionary 21d with the character patterns extracted by the character extraction unit 21c. To be done.

This character recognition block is, for example, FA
X It recognizes a character set in a predetermined position on the transmission original paper 21f and described in a character frame 21g in which a transmission destination is entered. Manuscript paper 21 with such a transmission address
When the transmission document is composed of a plurality of sheets, the f is used as the first (first) document. The image data read and input from the first original is stored in the image memory 21a for character recognition processing.

The area detection unit 21b is a predetermined FAX transmission document paper.
The position information of the character frame 21g is obtained from the format information of 21f, and the area in which the character to be recognized is described is detected. The character extraction unit 21c uses the area detection information and the information of the projection pattern of the image information, for example, the 24th
As shown in the figure, the image data of the characters described in the character frame 21g is individually extracted.

The identification unit 21e extracts the characteristics of the character pattern from the extracted character image and registers the extracted character pattern and the standard pattern dictionary 21d, as disclosed in, for example, Japanese Patent Publication No. Sho 49-12778. The pattern is matched with the standard pattern of each character. Then, the character category of the standard pattern that has been matched by this pattern matching is obtained as the recognition result.

Needless to say, the pattern matching method can be modified in various ways.

By the way, the second character recognition block for recognizing character information input online via a tablet or the like is, for example,
It is constructed as shown in Fig. 25.

The second character recognition block includes a coordinate detection circuit 21h that sequentially detects a series of position coordinates indicating a writing stroke of a character that is input online via a tablet or the like.

The time-series data of the position coordinates detected by the coordinate detection circuit 21h is input to the preprocessing circuit 21i, and the tablet 4
After a small noise such as a detection error in is removed, it is sequentially stored in the coordinate series storage circuit 21j and is used for character recognition processing. It should be noted that, for example, when a character for one character is input, the preprocessing circuit 21i performs a normalization process or the like on the size of the character.

In addition, the stroke number detection circuit 21k detects the stroke number of the character pattern, that is, the stroke number, for example, from the interruption of the writing stroke (time-series division of the position coordinate data).

Then, the recognition processing unit 21m selectively extracts the standard pattern of the corresponding stroke number from the standard patterns of the recognition target character category registered in the standard feature pattern memory 21n according to the stroke number information. Then, the characteristics of each stroke of the standard pattern and the characteristics of the stroke of the input character pattern stored in the coordinate series storage circuit 21j are compared with each other (matching processing). The answer determination circuit 21p determines the matching processing result, and obtains a recognition target character category having a stroke corresponding to the stroke feature of the input character pattern as the recognition result.

That is, according to the characteristic of the writing stroke of the character pattern input online, the characteristic of the stroke is matched with the characteristic of the stroke of the standard character pattern to recognize the input character pattern.

It should be noted that, as the feature of the stroke, it is sufficient to use the information of the position coordinates of the end point, the intersection, the folding point, etc. when the writing stroke is approximated by the broken line.

The character recognition unit 21 having the above-described functions character-recognizes character information read and input via a scanner or the like and character information input online via a position coordinate input device such as a tablet.

Next, the figure recognition unit 22 will be described.

The figure recognition unit 22 is configured, for example, as shown in FIG. The input unit 22a stores, for example, a captured and input graphic image, and provides the graphic image for recognition processing. The contour tracking unit 22b divides the tracking direction of the line segment into, for example, eight directions as shown in FIG. 27, and sequentially determines which direction the tracking direction is when the contour of the figure in the input image is tracked. There is. Specifically, for example, as shown in FIG. 28, a triangular figure is tracked in the clockwise direction, and information on the direction of the tracking is, for example, “1,2, ~ 2,3,4, ~ 4,5,7 , ~ 7 "is obtained as a series of direction codes.

The segmentation unit 22c extracts, for example, a singular point, such as a bent portion, from the direction code sequence thus obtained, and divides the contour of the figure into a plurality of characteristic portions according to the singular point. The matching unit 22d has information on the figure contour segmented in this way,
The input figure is recognized by performing a matching process with the characteristic information of various figures registered in the dictionary memory 22e.

For example, when the figure shown in FIG. 29 is given, for example, three contour points (i−1), (i),
At (i + 1), the sum of the direction codes is sequentially obtained, and this is smoothed as the direction code at the central contour point i. Noise components are removed by this smoothing process.

After that, the segmentation unit 22c detects an end point that is a characteristic point of the contour, that is, a point having a sharp bend, and divides the contour with the end point as the center. Then, each of the divided contour portions is collated with the dictionary memory 22e to obtain the recognition result.

As a result of the above processing, as illustrated in FIG. 30, the circle figure has no end points, the triangle figure has three end points detected, and the square figure has four end points detected.
Each of these figures is identified and recognized. At this time, information such as the fact that each of the above-mentioned end points is convex or that the contour connecting the end points is a straight line / curve may be used for figure identification.

On the other hand, the image recognition unit 23 is configured as follows.

FIG. 31 shows a schematic configuration of the image recognition unit 23, which is composed of an original image memory 23a, a binarizing device 23b, a processed image memory 23c, a thinning device 23d, and a code converting device 23e.

The image memory 23a stores the given image to be recognized, and the binarizing device 23b binarizes the image and stores it in the image memory 23c. This binarization level is
For example, the binarized image is variably set while monitoring the display.

Then, the thinning device 23d thins the binarized image image to form a line figure of the image. The image memory 23c is rewritten by the image image subjected to the line thinning processing and provided for the recognition processing.

The code converting device 23e is configured, for example, as shown in FIG. 32, and first, the segment dividing unit 23f divides the thinned image into a plurality of segments. This segment division is performed, for example, by dividing the line figure at its end points, branch points, or intersections. The curvature conversion unit 23g obtains the curvature of each of the plurality of segments thus divided.

Straight line / curve division unit 23h, curve division unit 23i, inflection point division unit 23j,
The inflection point dividing unit 23h further divides each segment divided as described above according to the information of the curvature thereof, and thereby, the inflection point, the switching point between the straight line and the curve, the inflection point, and the curve. A radius change point or the like is detected. By such segment division and feature point detection, information on each part constituting the image line figure is extracted.

Approximate information creation unit 23m, the information representing the image figure by integrating the information of these divided segments and the characteristic points in the segment, for example, the position coordinates of the start and end points of each segment, and the type of the segment Get code information that identifies

For example, when the input image image is given as shown in FIG. 33 (a), the image line figure 23n in the input image is thinned and extracted, and segmented as shown in FIG. In this example, an image line figure 23n in which a circular figure and a square figure are so-called skewered by a straight line is input. The image line graphic 23n is divided at its intersections as shown in FIG. 33 (b), and segmented into two semicircles, two U-shaped graphics, and four straight lines.

The curvature conversion unit 23g obtains the curvature of each segment divided as shown in FIG. 34, and the straight line / curve division unit 23h, the curve division unit 23i, the inflection point division unit 23j, and the inflection point division. The section 23h detects the characteristic point of each segment from the curvature change point. Specifically, in the example shown in FIG. 34 (a), the curvature at the refraction points of the two straight lines sharply increases, so that it is possible to detect the refraction point from this change in curvature. Further, in the example shown in FIG. 34 (b), since the change in the curvature is detected at the changing portion from the straight line to the curved line, the characteristic point can be detected from the change in the curvature.

Similarly, also in the example shown in FIGS. 34 (c) and (d), it becomes possible to detect the feature point in the segment from the change point of the curvature.

In this way, the image recognition unit 23 segments the given image figure and detects the characteristic points of each segment. Then, the image line figure is recognized by approximately expressing it as code information indicating each kind of a plurality of segments and its position coordinates.

Now, the voice collating unit 17 is configured as follows. The voice collating unit 17 is for performing individual recognition (personal identification) of the speaker who inputs the voice, and is configured as shown in FIG. 35, for example.

That is, the voice given through the voice input unit 17a is filtered by the phonological filter 17b and the personal filter 17c, respectively, and the voice feature thereof is extracted. Each band of the plurality of channels of the phonological filter 17b is set by equally dividing the audio frequency band as shown in FIG. 36 (a), for example. The phoneme filter 17b having such filter characteristics extracts the special parameter indicating the phoneme feature of the input voice. The bandwidth of each channel may be set by dividing the audio frequency band logarithmically.

On the other hand, the bandwidths of the plurality of channels of the personal filter 17c are set by exponentially dividing the audio frequency band as shown in FIG. 36 (b). With the personal filter 17c having such a filter characteristic, more voice features from the low frequency band to the middle frequency band of the input voice are extracted as compared with the features on the high frequency side.
The filter output of each of these channels is obtained as a characteristic parameter for personal verification.

Then, the word recognition unit 17d recognizes the word indicated by the input voice from the phoneme feature parameter obtained through the phoneme filter 17b by referring to the word dictionary 17e. The word recognition function is the same as that of the voice recognition unit 19 described above, and the function of the voice recognition unit 19 may be used as it is. Then, according to this word recognition result, a dictionary to be used for personal verification of the personal dictionary 17f is selected. In this personal dictionary 17f, the analysis results of the specific words uttered in advance by the individual subject to speaker verification by the personal filter 17c are classified and registered for each word.

Then, the speaker verification unit 17g calculates the similarity between each characteristic parameter of the corresponding word selected from the personal dictionary 17f and the characteristic parameter of the input voice obtained by the personal dictionary 17c, and the similarity is calculated. The values are discriminated by a predetermined threshold value. Then, by comparing the discrimination results with each other, for example, the personal category having the highest similarity value and the characteristic parameter having a sufficient difference with the next highest similarity value is regarded as the speaker of the input voice. Individually identified.

Here, the characteristics of the personal filter 17c will be described in more detail. As described above, the characteristics are different from those of the phonological feature filter 17b. Considering the distinctiveness of the individuality of the voice, the distinctiveness can be evaluated by an F ratio given as, for example, F ratio = (dispersion between individuals) / (dispersion within individual).

Now, when the F ratio of each channel output of the filter characteristic set in the phoneme filter 17b is examined, the exponential tendency shown by the solid line in FIG. 37 is shown. For this reason, conventionally, individual verification is performed exclusively by using the voice characteristic information on the high frequency side.

However, if it is possible to identify an individual by using the voice features of all frequency bands, it is considered that the matching accuracy is further improved rather than using only the features of the high frequency side of the voice. That is, if the value of the F ratio is 1 or more in all frequency bands and the inter-individual dispersion exceeds the intra-individual dispersion, it is possible to perform more accurate individual collation.

Therefore, here, as described above, the characteristic of the personal filter 17c is set exponentially, and the high-frequency side where the characteristic of individuality is remarkable is roughly extracted, and the number of channel allocation on the low-frequency side is increased. In this way, the voice feature on the low frequency side is finely extracted.

Specifically, since the change in the F ratio of each channel shows an exponential tendency, a filter bank in which the bandwidth of the high frequency side channel is exponentially increased as compared with the bandwidth of the low frequency side channel. And is used as a personal filter 17c.

According to each channel output of the filter 17c configured in this way, the F ratio is as shown by the broken line in FIG. 37,
A significant improvement in F ratio in the mid range is recognized. As a result, it is possible to perform personal verification by positively utilizing not only the voice characteristics of the high frequency side but also the voice characteristics of the mid frequency range.
It is possible to improve the matching accuracy.

That is, the voice collating unit 17 extracts the feature information that shows the individuality remarkably by devising the filter bank, in addition to the features used for word recognition of the input voice. As a result, individual identification of the speaker, that is, individual verification is performed with high accuracy independently of phonological recognition of the input voice.

Next, the voice synthesizer 26 will be described.

The speech synthesizer 26 includes a discriminator 26a and a decoder as shown in FIG.
26b, rule parameter generation device 26c, and speech synthesizer 26d
It is configured with.

The discriminator 26a determines whether the input code string is a character string or a code string indicating an analysis parameter for voice synthesis. This information determination is performed, for example, by determining the identification information added to the very beginning of the input code string. If it is determined to be an analysis parameter, the code string is given to the decoder 26b, and this is subjected to a decoding process to obtain the phoneme parameter and the prosody parameter, respectively.

When it is determined that the character string is the character string data, it is given to the rule synthesis parameter generation device 26c for the character string data and used for generation of the phoneme parameter and the prosody parameter.

The speech synthesizer 26d processes the sound source wave through the vocal tract approximation filter according to the phoneme parameter and the prosody parameter thus obtained by the decoder 26b or the rule synthesis parameter generation device 26c, and synthesizes the synthesized speech wave. Is generating.

Here, the rule synthesis parameter generation device 26c will be further described. The device 26c is configured as shown in FIG. The character string analysis unit 26e refers to the language dictionary 26 to identify each word in the input character string individually, and obtains accent information, word / bunsetsu boundary, grammatical information such as part-of-speech / inflection for the word. Then, the phonological rule and the prosody rule are applied to the analysis result, and the control information thereof is generated.

Here, the phonological rule provides information on the reading of the analyzed word, realizes phenomena such as confusion and unvoicedness caused by concatenation of words, and generates the phonological symbol string.
The voice parameter generation unit 26g inputs this phoneme symbol string,
According to the syllable unit, syllable parameters are sequentially obtained from the CV file 26h and interpolated and combined. The phoneme parameter generating unit 26g generates a phoneme parameter sequence from the phoneme symbol string.

The prosody rule is to determine the boundaries of utterances and breathing positions according to grammatical information such as word / bunsetsu boundaries, and to determine the duration and pause length of each sound. At the same time, this prosodic rule generates a prosodic symbol string based on the basic accent of each word and considering the phrase accent. The prosody parameter generation unit 26i inputs this prosody symbol string and generates a prosody parameter string representing the time change pattern of the pitch.

On the other hand, when the input code string is the code string indicating the analysis parameter for speech synthesis, the decoder 26b functions as follows.

That is, when the code string of the analysis parameter indicates the cepstrum coefficient of the CV file, the code string 26m generally corresponds to the parameters P (pitch) and C ₀ , C ₁ , ... Cm (cepstral coefficient) as shown in FIG. Information is compressed by bit allocation. Therefore, the decoder 26b uses the parameter conversion table 26n to convert / decode the information-compressed analysis parameter into the number of bits suitable for the speech synthesizer 26d. For example, each parameter is converted to 8 bits,
A phonological parameter sequence (cepstral coefficient) and its prosodic parameter sequence (pitch) are obtained.

The voice synthesizer 26d uses the voice source 2 as shown in FIG. 41, for example.
6q and an unvoiced sound source (M sequence generator) 26r are provided to selectively generate a voiced sound source wave (P ≠ 0) or an unvoiced sound source wave (P = 0) according to the pitch data P of the input prosody parameter sequence. ing. This source wave is input to the preamplifier 26c, level-controlled according to the cepstrum coefficient C ₀ of the phoneme parameter, and the logarithmic amplitude approximation digital filter 26t.
Entered in. This logarithmic amplitude approximation digital filter 26
Reference numeral t denotes a resonance circuit that approximates the vocal tract characteristics according to the cepstrum coefficients C ₁ to Cm of the phoneme parameters, and filters the sound source wave. The logarithmic amplitude approximation digital filter 26t synthesizes and outputs the voice data represented by the phoneme parameter and the prosody parameter.

Then, the signal synthesized by the logarithmic amplitude approximation digital filter 26t is passed through the D / A converter 26u, filtered through the LPF 26v, and output as a synthesized voice signal (analog signal).

In the speech synthesis unit 26 configured as described above, the speech represented by the data series is regularly synthesized and output from the input data series.

Next, the image composition unit 27 will be described.

The image synthesizing unit 27, as shown in FIG.
a, display file memory 27b, image synthesis circuit 27
c, image memory 27d, and optional display
Composed of 27e. This display 27e is
It may be the display unit 10 prepared for the workstation.

The image synthesizing circuit 27 reads the vector, polygon, and arc parameters written in the display file 27b under the control of the dedicated control calculation unit 27a, generates the line figure indicated thereby, and generates the image memory. Writing to the specified address in 27d. This image synthesis circuit
The specified line figure image is constructed on the image memory 27d by the image generation function 27. The line graphic image is displayed on the display 27e and monitored under the control of the control calculation unit 27a.

The image generating circuit 27b has a special processing function for image generation and a painting processing function. This special processing function is, for example, a function of erasing hidden lines or clipping processing for overlapping of a plurality of image figures. The filling function is a process of filling a partial area of the image figure with a designated color.

By the function of the image synthesizing circuit 27b as described above, various image figures are created, and their synthesizing processing and the like are performed.

By the way, the synthesis of the image figure and the natural image generated as described above is roughly classified into the following two. One is a process of embedding and synthesizing an image image obtained by arithmetic processing in a natural image such as a landscape photograph as a background, and the other one is that the control arithmetic unit 27a has as an internal model. It consists of the process of embedding and synthesizing a natural image in a certain plane image.

Here, when embedding an image image in the former natural image, a code indicating "transparent color" is given to a graphic generated by the control calculation unit 27a as illustrated in FIG. 43, for example. It can be achieved by superimposing this on a natural image and synthesizing it. Then "transparent color"
In the image area to which the code is given, the information of the natural image is displayed as it is.
The figure in which a occurred is displayed. As a result,
Image composition with a natural image as a background will be realized. This technique is called overlay.

On the other hand, as shown in the concept in FIG. 44, a natural image may be written in the image memory, and the graphic generated by the control calculation unit 27a may be written on (near) the natural image. . This method is called the z-buffer method and can be realized relatively easily together with the above-mentioned overlay method.

By the way, in the latter case in which a natural image is inserted and combined in a plane shown as an internal model of the control calculation unit 27a, high speed processing is performed as follows.

The coordinate transformation necessary for embedding a natural image on a plane in a plane facing an arbitrary direction in a three-dimensional space is given by the following equation.

However, X and Y are coordinates on the display surface, and u and v are coordinates on the natural image.

If this coordinate conversion process is executed as it is, 6 times of multiplication and 2 times of division are required every time one pixel is displayed, which requires a huge amount of calculation and a large amount of calculation processing time.

Therefore, here, the above-described calculation is divided into two conversion processes and executed via the following intermediate coordinates (s, t). This arithmetic processing is executed at high speed using, for example, affine transformation.

u = (α ₁ s + α ₂ t + α ₃ ) / t (1) v = (α ₇ s + α ₈ t + α ₉ ) / t s = C ₅ X−C ₄ Y (2) t = C ₄ X + C ₅ Y + C ₆ That is, the above The perspective transformation is performed using the equation (1), and then the two-dimensional affine transformation is performed using the equation (2) to perform the perspective transformation to an arbitrary plane at high speed.

Here, since the denominator of the equation (1) is the coordinate t itself, it is easy to execute the calculation at high speed by slightly improving the conventionally known affine transformation circuit.

In this way, the image synthesizing unit 27 executes various image synthesizing processes at high speed.

Next, the database unit 32 will be described.

The database unit 32 organizes and stores various types of data such as codes, images, and voices, and supplies this to various application systems. FIG. 45 shows a schematic configuration of the database unit 32. As an interface unit 32a for executing a command analyzing process and the like, a data operating unit 32b for executing a database searching process, etc., as a storage medium for storing various data. Magnetic disk device 32c and optical disk device 3
2d and its additional function part 32c.

Various data are classified and organized into a plurality of relations according to the type of the data, and each relation is registered to construct a database.

Hereinafter, the database unit 32 will be described by dividing it into four, that is, its logical structure, stored data, physical structure, and additional function.

The logical structure indicates how various data are stored when the database unit 32 is viewed from the application system side. Here, the data is handled as a logical structure according to the relational model, for example, as an image of a table as shown in FIG.

Some columns (attributes) in the table (relation)
Is provided, and data in a predetermined unit is stored in each of these columns. The unit of data (tupple) is
It is defined as a set of values to be stored in each column. One relation is constructed by an arbitrary number of attributes storing such tuples.

In this model, however, data is stored in the database by specifying the relation name and giving the value of each attribute. Also, the database search specifies relations and attributes and determines whether the value stored therein meets the specified condition with the specified value or the value stored in another attribute. Then, a tuple that satisfies the condition is extracted.

The search conditions are given as their values being equal, not equal, small, large, etc. At this time, it is also possible to specify a search condition for each of the plurality of attributes and perform a logical process (AND, OR, etc.) on the result of the condition determination. Furthermore, a database search that finds a predetermined tuple from among multiple relations by specifying multiple relations and the condition that the value of a certain attribute of a certain relation is equal to the value of a certain attribute of another relation. Is also possible.

Further, the data deletion from the database is basically performed in the same manner as the above search, but instead of extracting the tuple, the tuple is deleted.

Further, the data update is similar, and is performed by changing the value of the specified attribute of the obtained tuple and storing it.

In addition, information (personal name and person-in-charge code) of a person who is permitted to read, add, and change data is written in each relation for each attribute, and measures for data protection are taken. It is also possible to take this data protection measure in relation units instead of taking it for each attribute. There may be a plurality of pieces of person information described here.

In the relation example shown in FIG. 46, the data is shown as a character string, but the data accumulated in each relation may be a simple bit string. In other words, the data stored in the relation may be not only a character string but also image information, voice information, or the like.

Now, the data accumulated in this database includes, for example, the relations of the "personal schedule" shown in Fig. 46 described above, such as "address book""personal work and its agents""operation" as shown in Fig. 47. "History""Personnel""Meetingroom"
It consists of various relations such as "meeting room reservation" and "meeting".

As shown in this example, relations are mainly used for personal use and commonly used by many users. And a personal relation is provided for each workstation used by each individual,
Also, the common relation is provided in a workstation common to a plurality of users.

The common workstation does not necessarily mean that its hardware is different from other workstations. Further, it goes without saying that a personal workstation may also serve as a common workstation.
Further, the common workstation is not limited to one, and a plurality of common workstations may be provided depending on the hierarchical level of the system. In short, a common workstation is set as one that can be easily specified from a plurality of workstations.

Here, the data structure of the "personal schedule" relation shown in FIG. 46 will be briefly described.

This relation indicates that the relation name is “personal schedule” and is created by “ΔΔΔΔ”. This relation creator "△△△
“A” allows all data operations for the relation.

In addition, according to the data protection function added to this relation, reading of data is permitted to all, and addition of data is permitted only to "○○○○" and "persons who belong to the engineering department". ing. The “person who belongs to the engineering department” is obtained by referring to the relation of “personnel”, for example. In addition, data change is allowed only for those whose "person level" value is "5" or more. This "person level" is related to personnel relations, for example (manager; 8) (deputy manager; 7) (section manager; 6) (chief manager; 5)
Represents a position such as.

Furthermore, for this relation, "start time""endtime"
Attributes such as “type”, “name”, and “place” are set, and data is written in each of them.

Next, a physical structure for actually storing the above-mentioned various relations in the database unit 32 will be described.

The information storage unit (storage unit) stores a large amount of data, and can read and write any part thereof at a relatively high speed, and the magnetic disk device 32c and the optical disk device 32g described above are used as being inexpensive in terms of price. .

The storage of the database in the information storage unit has a storage area of the information storage unit of a specific size (for example, about several kilobytes, which is determined according to the tapple length, the speed of the computer, etc.).
Each page is divided and managed as a page. Then, as shown in FIG. 48, for example, page management information is stored in the 0th page, relation list information is stored in the 1st page, and page information in use is stored in the 2nd page.

This list of relations shows where the various relations are in the database.

For example, the actual data stored on the 9th and 11th pages may be sorted according to the information of the index page stored on the 10th page based on the relation attribute (main attribute) stored on the 5th page. Has become. The information on this index page is
It indicates on which page several or several attribute values are stored.

If you want to search data by an attribute other than this main attribute, please refer to that attribute.
First, the sub data shown on the 21st or 22nd page is obtained via the sub index of the page. This sub data contains only the value of the attribute and the value of the above-mentioned main attribute, and the actual data is obtained using the value of the attribute obtained here.

If the amount of actual data such as image data or audio data is enormous and some bit errors in it are not a problem, these actual data can be transferred to another inexpensive optical disk device 32d or the like. You may make it file in an information storage device. In this case, the fact and the storage position information of the actual data in the device may be stored in the actual data pages such as the ninth page and the eleventh page.

However, the additional function to the database thus constructed consists of, for example, automatic discarding of unnecessary data. This automatic discarding of unnecessary data is performed by giving [discard; enable / disable] [discard method] and the like as additional information of the relation and operating an erase command for each relation at a predetermined interval.

Note that the tapple can be deleted by determining whether or not the ending time of the conference information is before the current time, for example. Therefore, it is not necessary to add a special function to erase such a tuple.

In addition, there is data preservation as another important function of the additional function. This data security function prevents the data from being illegal (randomized or lost) due to, for example, a hardware failure or a power failure. Specifically, this data security function is realized by duplication of information and writing on a magnetic tape.

In this way, the database unit 32 classifies and organizes various data for each relation and manages them in page units to provide them to various application systems.

Next, the work environment data collection unit 25 will be described.

The work environment data collection unit 25 collects past operation history data for the workstation and provides operation guidance based on the collected data.

Here, in the work environment data collection unit 25, for example, as shown in FIG. 49, there is a command correspondence table that associates commands corresponding to functions of the information processing system with commands corresponding to functions of other information systems. It is provided.

Specifically, when the information processing system is A and the other information processing systems are B, C, D ..., the command of the other system corresponding to the command “DELETE” in the system A is “D”.
It is indicated by the command correspondence table that the EL is "ERASE" and "REMOVE".

FIG. 50 shows a schematic configuration of a work environment data collection unit 25 that analyzes a command input by the user and executes a predetermined operation and various guidance.

In this work environment data collection unit 25, first, the command input unit
The command input from 25a is given to the command analysis unit 25b and analyzed by referring to the command correspondence table 25c. Specifically, it is checked whether the input command is registered in the command correspondence table 25c according to the procedure flow shown in FIG. That is, when a command is input, it is first checked whether or not the input command belongs to the system A. Then, when the input command is analyzed as a command of the system A, the command analysis section 25b gives the input command to the command execution section 25d and causes it to execute a predetermined operation based on the command.

On the other hand, if the input command does not belong to the system A, it is checked whether or not a command from another system is applicable, and if there is a corresponding command,
The corresponding command is displayed on the screen display unit 25e. That is, in the case of a command used in another system (system B), for example, "DEL", the corresponding command "DELETE" of system A is obtained and displayed on the screen display unit 25e as operation guidance. Will be done.

If the command corresponding to the input command does not exist in the command correspondence table 25c, a command error message is displayed on the screen display unit 25e.

Specifically, the processing for the command input is performed as follows. Now, it is assumed that a user who has experience operating the systems B and C operates the system A (information processing system) for the first time. Here, when the user inputs a command to delete the data "ABC", conventionally, it is necessary to search for the command "DELETE" for deleting the data according to the instruction manual of system A and input it. .

However, here, according to past experience, the user, for example, uses the data erasure command “ERAS
Enter "E ABC" as shown in Fig. 52 (a).

Then, the work environment data collection unit 25 analyzes this input command, and from the command correspondence table 25c, the command "DELETE" of the system A corresponding to the input command "ERACE".
Will be displayed as a guide. As a result, even when the user operates the system A for the first time, he / she can know that the data erasing command is “DELETE” and enter the command according to the guide to erase the data. It will be possible.

Also, in order to display a list of file names, FIG. 52 (b)
When the command "DIR" in the system B is input as shown in Fig. 3, the corresponding command "CATA" in the system A is similarly obtained and displayed as a guide. As a result, by entering the command "CATA" according to this guide, a list of its filenames will be displayed.

In this way, by utilizing the function of the work environment data collection unit 25, the corresponding command in the system is guided and displayed by the input of the command used in the system having the past operation experience. Therefore, the system user can operate the system by making the most of the knowledge acquired in the past. Then, it becomes possible to easily know the command of the information processing system. Therefore, it is not necessary to check the operation manual of the information processing system each time. Therefore, it is expected that the time required for learning the operation of the system can be significantly reduced.

Alternatively, a command corresponding to the input command may be obtained, and when this is displayed as a guide, the command may be executed in response to the determination input of the pass / fail.

That is, FIG. 51 shows the flow of the procedure, and FIG. 54 shows a display example thereof.
E ", and the corresponding delete command“ DELE ”of system A
When "TE" is requested, it is inquired whether or not this is correct. When there is a correct (Y) instruction input, it is determined that the input command indicates "DELETE", and this is executed by the command execution unit 25d. Send it to and let it do its work.

In this way, the desired processing is executed in accordance with the input command at the same time that the command correspondence is instructed, and it is not necessary to input the correct command again. That is, the input command is automatically converted into the corresponding command, and the processing is executed. Therefore, it is possible to further improve the operability.

It should be noted that there may be any number of corresponding commands depending on the type of system. In short, command correspondence table 25c
It may be stored in association with. It goes without saying that the command is not limited to the character string format described above.

Next, data collection of the system proficiency level in the work environment data collection unit 25 will be described. An external storage device and a control device are provided inside the work environment data collection unit 25 as hardware for performing the data collection process of the system proficiency level.

FIG. 55 is a flowchart showing the data collection process of the system proficiency level.

User's identification code (user number, password, etc.)
When is input, the work environment data collection unit 25 obtains a proficiency level table corresponding to the identification code from the external storage device and sets it in the device. This proficiency level table stores the level of proficiency of each user with respect to various functions of the system, and is configured as shown in FIG. 56, for example.

That is, this proficiency level table shows the usage frequency of each function, the last usage date, the proficiency class for the function declared by the user, the proficiency class when the function was last used, and the proficiency level of the function. It is composed of complexity information and the like.

Here, the complexity is higher as the applicable function requires specialized knowledge and becomes higher as the function is higher than the basic function.

However, such a proficiency table is provided for each user,
Each is stored in an external storage device. For the user who uses the system for the first time, a proficiency table for the user is created by newly setting the identification code and registered in the external storage device.

Incidentally, in the external storage device, as shown in FIG. 57, for example, in addition to the proficiency level table described above, a message for each used function corresponding to the proficiency level class is registered. The lower the proficiency level of this message, the more clear the explanation including the background explanation. Also, the higher the level of proficiency,
It has a high level of content including a brief explanation and a professional introduction of the function.

The proficiency classes are classified and set as, for example, A; beginner class B; intermediate class C; proficient class.

Then, when the proficiency level table corresponding to the input identification code is requested, a menu for allowing the user to select the function to be used is displayed next. For this menu, the user inputs, for example, a number corresponding to the used function.
Then, the control device determines whether the input information is the end signal or the selection signal of the function to be used, and in the case of the function selection signal to be used, operates as follows.

That is, when the use function selection signal is input, first, referring to the proficiency table for the user, information such as the use frequency, the last use date and time, the declared proficiency level, etc. corresponding to the selected use function is displayed. Desired. Then, weighting processing is performed according to these pieces of information, and the current proficiency level class is determined.

This proficiency class is determined, for example, by using the usage frequency Pi, the last usage date and time Te, the current usage date and time Tc, the user-reported proficiency class X ₁ , and the previous usage proficiency class X ₂ ε. {A,
B, C}, the complexity is Pc, and the discriminant function is Fr, Fr = K ₁ Pi + K ₂ (Tc-Te) + K ₃ G ₁ [X ₁ ] + K ₄ G ₂ [X ₂ ] + K ₅ Pc Sought. However, in the above equation, K ₁ , K ₁ , K ₃ , and K ₄ are
It is a constant that is set to an appropriate value through experiments and the like. Also, G ₁ and G ₂ above are And Y ₁ , Y ₂ , Y ₃ , Z ₁ , Z ₂ , and Z ₃ are evaluation weights for A, B, and C. These evaluation weights have a relationship of Y ₁ <Y ₂ <Y ₃ and Z ₁ <Z ₂ <Z _{3 and} are set to appropriate values by experiments or the like.

Here, G ₁ [X ₁ ] takes a value Y ₁ when X ₁ = A, and X ₂ =
In the case of B, it means to take the value Y ₂ . Also (Tc-T
e) is a time conversion of the number of days from the date of last use to the present.

Then, the class determination is performed as follows based on the value of the discriminant function Fr described above.

Fr <N ₁ ... A class N ₁ ≦ Fr <N ₂ ... B class N ₂ ≦ Fr ... C class The determination thresholds N ₁ and N ₂ are appropriately determined based on experiments and the like.

When the proficiency level is determined in this way, a guide message or an error message corresponding to the determined proficiency level and corresponding to the designated use function as described above is obtained from the external storage device.

Thereafter, the proficiency class determined this time and the previous proficiency class stored in the proficiency table are compared and collated. When the proficiency level is changed, a message indicating that the proficiency level is changed is added to the guide message and the like.

The proficiency class change message is composed of, for example, four types of messages as shown in FIG. Then, it is determined according to the mode of class change and is displayed together with the guide message and the like. The user performs the processing operation according to the various messages displayed in this way.

Specifically, for the use function of storing the created data in a file, when the user is judged to be a beginner class (class A), a message as shown in FIG. 59 is displayed. If the user makes a mistake in inputting the information despite this message, an error message as shown in FIG. 60, for example, is displayed to guide the operation of the function to be used.

If the user's proficiency level is determined to be the intermediate class (B class), a message as shown in FIG. 61 is displayed. If the user makes a mistake in inputting information despite this message, an error message as shown in, for example, FIG. 62 is displayed to guide the operation of the function to be used. Similarly, if it is determined that the user's proficiency level is the proficiency class (C class), the 63rd
When a message as shown in the figure is displayed and there is an error in the information input, for example, an error message as shown in FIG. 64 is displayed to guide the operation of the function to be used.

However, if data is entered in the blank space of the guide message displayed as described above, the control device increments (+1) the corresponding usage frequency in the proficiency level table of the corresponding user, which is obtained as described above, and makes a final use. Update the date, time, and previous learning class. Then, the execution of the used function is urged, and it is considered that the corresponding used function has ended, and the process returns to the menu display operation for selecting the used function described above.

Here, when the use function selection signal is input again, the above-mentioned processing is repeated and executed again. However, when the end selection signal is input, the proficiency level table created / updated as described above is written in the proficiency level file of the external storage device together with the identification code of the corresponding user, and the proficiency level table is saved. Then, the series of processing procedures is ended.

In this way, the work environment data collection unit 25 collects data of the proficiency level regarding the operation of the system, and appropriately guides the operation according to the collected data.

The above is the basic configuration of this workstation and its functions.

Next, a media conversion method which is a feature of the present invention will be described.

The media conversion performed here includes, for example, recognizing the input voice by the voice recognition unit 19 and converting it into a character code string, or synthesizing the character code string by voice synthesizing process by the voice synthesizing unit 26. This is a process of generating voice. In addition to conversion between this voice media and character code media, this workstation has various media conversion functions such as figure recognition / figure synthesis, image recognition / image synthesis, and compression / decompression processing of voice and images. Is equipped with.

A feature of the present invention is a function of selectively controlling such media conversion according to the type of a communication partner terminal for information communication, converting the media into a medium that the communication partner terminal can process, and outputting the communication. Mainly output form selection unit 24
Will be held in.

The output form selection unit 24 is activated in response to a media selection request signal from the control unit 2 and selects which medium is used to output data. That is, it is used to select which medium, out of various media such as a character code string, voice, and image, is used to output information through communication.

FIG. 65 is a schematic configuration diagram of the output form selection unit 24, which is provided with a media selection control unit 24a, an input media determination unit 24b, a partner media determination unit 24c, a media conversion table 24d, and a self media function table 24e. To be done. Also 66th
The figure shows an outline of the processing flow of the output form selection unit 24. The function of the output form selection unit 24 will be described along the flow of this processing procedure.

When the media selection request signal is given, the media selection control unit 24a requests the control unit 2 to provide the input media information necessary for the media selection operation. Then, it issues a media information detection request and a media function identification request to the input media determination unit 24b.

The input media determination unit 24b includes a media detection unit 24f and a media identification unit 24g. Then, in response to the information request from the medium selection control section 24a, the information of the input medium given from the control section 2 is detected, and the function of the detected medium is discriminated and determined. By this identification determination, for example, the medium of information input to communicate with various information terminals is identified and detected. Specifically, for example, when audio data is given, the input media determination unit 24b identifies and determines that the input media is voice and the function of the media is ADPCM.

After that, the media selection control unit 24a sends a data output destination to the control unit 2 to another functional block of its own terminal (in the work station) or to another unit connected via a communication line or the like. Inquire whether it is a workstation or communication terminal. Then, when instructed to output data to another workstation or communication terminal,
The media selection control unit 24a requests the control unit 2 for identification information regarding the transmission partner station. The information about the partner station that receives this request and outputs data is the partner media determination unit 24c.
Entered in.

The partner media determination unit 24c includes a partner station identification unit 24h, a partner station media identification unit 24i, and a function identification unit 24j, and operates in response to an identification information determination request from the media selection control unit 24a. Then, the type of the partner station is first identified from the identification information about the partner station, and the media of the partner station is identified. Then, the function of the partner station media is identified.

Specifically, according to the information given from the control unit 2, for example, the communication partner station that outputs (transmits) data is an automatic FAX, its communication medium is an image, and its function is
Identify GIII type, etc. The identification of the partner station may be performed based on the information directly sent from the partner station using the negotiation (handshake) function. If the communication partner station does not have the negotiation function, its media detection function is
It should be held in 24j. This makes it possible to identify the function according to the media information signal from the other party.

FIG. 67 shows the flow of the identification processing procedure of this partner station. As shown in this flow, the identification processing is performed, for example, by first determining whether or not the communication partner station is a telephone terminal, and if it is a telephone terminal, determining whether or not a FAX signal comes from the telephone terminal. Be seen.

Then, when the partner station is a telephone and a FAX signal arrives, it may be identified that the partner device is a FAX. If it is determined that the telephone is a telephone and the FAX signal does not arrive, it may be determined that the partner device is a normal telephone. Further, when it is determined that it is not a telephone, it is possible to determine that the partner device is a communication device other than the telephone.
Through such a determination process, the type of the communication partner station is determined, and the type of media that the communication partner station can process is determined.

Then, when the media of the communication partner station is identified and determined, the media selection control unit 24a then refers to the media conversion table 24d configured as shown in FIG. 68, for example, to input media, input function, and partner. Device, partner device media,
Obtain media conversion selection information corresponding to the function of the partner device. That is, when the input information is transmitted to the communication partner station, information is obtained as to how the input information should be converted into media and which media conversion means is necessary.

For example, if the input medium for information communication is voice, the function is ADPCM, and the partner device is a GIII type fax, the media that can be processed by the partner device is an image. It is required to execute (voice) to (code character) (code character) to (image). At the same time, we demand that the conversion function be realized by (ADPCM; voice) to (GIII; FAX). At this time, if subordinate media conversion information exists, this is also obtained at the same time.

The media conversion information obtained in this way is the control unit 2
And the format of the data output is selectively designated. As a result, the voice data input in the ADPCM data format is once converted into a character code string by the voice recognition unit 19 and further converted into a character image by the image synthesis unit 27. Since the communication partner station is FAX, it is converted into a GIII mode signal (communication medium) and transmitted and output from the communication devices 12 and 13. By such selection control of the media conversion function, the information input by voice is communicated and output to the communication partner terminal having the image data output function.

When data is output to the inside of the workstation of its own, the media selection control unit 24a refers to the own media function table 24e to obtain an output format capable of outputting data. In accordance with this information, the media selection control unit 24a refers to the self media conversion table of the media conversion table 24d, similarly obtains the media conversion information, and gives this to the control unit 2.

According to the media conversion information obtained in this way,
For example, the sentence synthesizing unit 26 described above is used for media conversion of sentence information given as a character code sequence into voice information, or the voice recognition unit 19 is used for media conversion of voice information into character code sequence information.

If the communication partner terminal is compatible with a plurality of types of media, one of the media may be selected to selectively control the media conversion. At this time, if the selected partner medium is in use, a free medium may be searched for and the selection may be controlled.

Further, when the communication partner terminal inquires of the workstation about the type of media that the workstation can handle, it responds to the inquiry by referring to the self media function table 24e, The type of station or the type of media that can be output by the workstation may be notified. Moreover,
The communication medium may be selected in consultation with the communication partner terminal.

Alternatively, the selection control for media conversion may be performed by determining whether the communication partner terminal is set to the automatic mode for media selection or the manual mode.

[Effects of the Invention] As described above, according to the present invention, an input medium of information to be communicated is detected, a medium that can be processed by a communication partner terminal is detected, and a media conversion mode is selected according to the detection result. Therefore, even if the function of the communication terminal is not the same as that of the workstation, information communication can be effectively performed with the communication partner terminal. Therefore, it becomes possible to communicate and output information to various types of terminals by using media that can be handled by the terminals, and it is possible to expand the communication range and convenience for workstation users. In addition, a great effect in practical use can be obtained such that information communication can be performed by effectively using various media.

[Brief description of drawings]

FIG. 1 shows an embodiment of the present invention. FIG. 1 is a diagram showing characteristic processing functions of a workstation according to the present invention, FIG. 2 is a schematic configuration diagram of the workstation, and FIG. 3 is a workstation. 4 is an external view of the IC card attached to the IC card, FIG. 4 is an exploded perspective view showing the structure of the IC card, FIG. 5 is a view showing the structure of the printed circuit board portion of the IC card, and FIG.
FIG. 7 is a diagram showing a configuration of a semiconductor integrated circuit unit of an IC card, FIG. 7 is a diagram showing a configuration of an encryption processing unit in a workstation, FIG. 8 is a diagram showing a concept of encryption / decryption, and FIG. 9 is encryption. Fig. 10 is a block diagram of the section, Fig. 10 is a block diagram of the decoding section,
FIG. 11 is a block diagram of the RSA processing unit, FIG. 12 is a diagram showing the configuration of the image matching unit in the workstation, FIG. 13 is a diagram showing an example of a face subjected to image processing, and FIG. 14 is a diagram showing image data. FIG. 15 is a diagram showing a configuration, FIG. 15 is a diagram showing a configuration of a voice recognition unit in a workstation, and FIG. 16 is a diagram showing an example of an input voice pattern.
FIG. 17 is a diagram showing acoustic characteristics of consonants, FIG. 18 is a diagram showing an example of a transition network, FIG. 19 is a diagram showing a procedure of a voice recognition process, and FIG. 20 is a description of partial segment detection for an input voice. FIG. 21 is a diagram showing a learning processing procedure of the voice recognition dictionary, and FIG. 22 is a first part of the character recognition unit in the workstation.
FIG. 23 is a diagram showing the configuration of a character recognition block of FIG. 23, FIG. 23 is a diagram showing an example of a FAX transmission original sheet in which characters to be recognized are described, and FIG. FIG. 25 is a diagram showing the configuration of the second character recognition block in the character recognition unit, FIG. 26 is a diagram showing the configuration of the graphic recognition unit in the workstation, and FIGS. 27 to 30 are for explaining the graphic recognition processing. FIG. 31 is a diagram showing the configuration of the image recognition unit in the workstation, FIG. 32 is a configuration diagram of the code conversion device, and FIG.
Fig. 34 is a diagram showing an example of processing for an input image. Fig. 34 is a diagram showing feature point detection in a segment. Fig. 35 is a diagram showing the structure of a voice collating unit in a workstation. Fig. 36 is a band division example of a filter bank. FIG. 37 is a diagram showing a filter characteristic, FIG. 38 is a diagram showing a configuration of a voice synthesis unit in a workstation, FIG. 39 is a configuration diagram of a rule synthesis parameter generation device, and FIG. 40 is a voice parameter. FIG. 41 is a diagram showing a conversion structure.
Fig. 42 is a block diagram of a speech synthesizer, Fig. 42 is a diagram showing the configuration of an image synthesis unit in a workstation, Figs. 43 and 44 are diagrams showing the concept of image synthesis processing, and Fig. 45 is a database in a workstation. Fig. 46 shows the structure of parts, Fig. 46 shows the data structure of the database, Fig. 47 shows an example of relations, Fig. 48 shows the structure of relations, and Fig. 49 shows the structure of command correspondence table. FIG. 50 is a diagram showing the configuration of the work environment data collection unit in the workstation, FIGS. 51 to 54 are diagrams for explaining the processing of the command unit, and FIG. 55 is system proficiency level data. FIG. 56 is a diagram showing the flow of the collection process, FIG. 56 is a diagram showing the structure of the proficiency level table, FIGS. 57 to 64 are diagrams for explaining the process of the work environment data collection unit, and FIG. 65 is the present invention. Select control of media conversion FIG. 66 is a diagram showing the configuration of the output form selection unit, FIG. 66 is a diagram showing the flow of the output form selection processing procedure, FIG. 67 is a diagram showing the flow of the partner station identification processing procedure, and FIG. 68 is the structure of the media conversion table. FIG. 1 ... bus, 2 ... control unit, 3 ... image input device,
4 ... Position input device, 5 ... Voice input part, 6 ... Keyboard part, 7 ... IC card part, 8 ... Bus controller,
9 ... Voice output device, 10 ... Display unit, 11 ... Image output device, 12, 13 ... Communication device, 14 ... Switching device, 15 ... Timer unit, 16 ... Encryption processing unit, 17 ... … Voice verification unit, 18 …… Image verification unit, 19 …… Voice recognition unit,
20 …… Speech analysis section, 21 …… Character recognition section, 22 …… Figure recognition section, 23 …… Image recognition section, 24 …… Output form selection section, 25
…… Work environment data collection unit, 26 …… Speech synthesis unit, 27 ……
Image synthesizing unit, 28 ... Graphic synthesizing unit, 29 ... Voice compression / decompression unit, 30 ... Image compression / decompression unit, 31 ... Signal processing unit, 32 ... Database unit.

Front page continuation (72) Inventor Hiromi Saito 1 Komukai Toshiba-cho, Sachi-ku, Kawasaki-shi, Kanagawa Inside the Toshiba Research Institute Co., Ltd. (72) Inventor Makoto Negishi Komukai Toshiba-cho, Kawasaki-shi, Kanagawa (56) Reference JP-A-58-194454 (JP, A) JP-A-57-185535 (JP, A) JP-A-61-187451 (JP, A)

Claims (6)

[Claims] 1. A multimedia-compatible workstation comprising a plurality of types of media conversion means, the workstation comprising an input media detection unit for detecting an input media of information used for communication, and a communication partner terminal. A partner media detection unit that detects a processable media and obtains a communication medium for communicating the information, and a unit that refers to a media conversion table and obtains the type of media conversion according to the input media and the communication medium. A media conversion system comprising: a media conversion means for selectively activating the media conversion means in accordance with a required type of media conversion, converting the input information into a media, and transmitting the media to the communication partner terminal. . 2. The media converting means comprises a voice recognition device, a character recognition device, a figure recognition device, an image recognition device, a voice synthesis device, an image synthesis device, a graphic synthesis device, and a compression / expansion device for various data. Claim 1 that is selectively used according to the type of media conversion
Media conversion method described in section. 3. The media conversion method according to claim 1, wherein the partner media determination unit detects the type of the communication partner terminal and obtains the type of media that can be processed by the communication partner terminal. 4. The media conversion according to claim 1, wherein the partner media determination unit determines the communication media by selecting one of a plurality of media that can be processed by the communication partner terminal. method. 5. The partner media determination unit determines whether or not the communication partner terminal automatically sets the processing media according to the communication media, and determines a communication media suitable for communicating information. The media conversion method described in the first item of the range. 6. The partner media determination unit is provided with a function of notifying the communication partner terminal of the type of the workstation or the type of media that can be output by the workstation in response to a request from the communication partner terminal. The media conversion system according to claim 1.