专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明专利

JP2000163405A DATA DECOMPOSING DEVICE WITH EVENT CUTTING MECHANISM 失效
公开(公告)号：JP2000163405A
公开(公告)日：2000-06-16
申请号：JP33608698
申请日：1998-11-26
申请人： FUJITSU LTD
发明人： YAMAKAWA HIROSHI , OKADA HIROYUKI , WATABE NOBUO
IPC分类号： G06F17/18
摘要： PROBLEM TO BE SOLVED: To provide a data decomposing device including the event cutting mechanism which can select a candidate state composed of segments including a small number of events. SOLUTION: The data decomposing device which extracts a partial state whose evaluation quantity has an extremal value in both a feature quantity selection direction and a partial event selection direction from data having a matrix structure of feature quantities and events is provided with the event cutting mechanism. The event cutting mechanism selects events in order from a segment including more events and searches for the extremal value of evaluation of all segments including more events to extract the candidate state. In this case, when the 1st extremal value is found, events included in the segment which is selected at this time are cut and a similar process is repeated for segments of the remaining events to extract the partial state.

2. 发明专利

JPH11104984A REAL ENVIRONMENT INFORMATION DISPLAY DEVICE AND RECORDING MEDIUM IN WHICH PROGRAM FOR EXECUTING REAL ENVIRONMENT INFORMATION DISPLAY PROCESS IS RECORDED AND WHICH CAN BE READ BY COMPUTER 失效
公开(公告)号：JPH11104984A
公开(公告)日：1999-04-20
申请号：JP27242097
申请日：1997-10-06
申请人： FUJITSU LTD
发明人： NAITO HIROHISA , SEKIGUCHI MINORU , OKABE HIROYUKI , WATABE NOBUO
IPC分类号： B25J13/00 , G05D1/02 , G06F17/00 , G06F19/00 , G06T1/00 , G06T7/00
摘要： PROBLEM TO BE SOLVED: To enable a user to pertinently and easily grasp the state of real environment by displaying a three dimensional image of the real environment in which the visual point, time, range, etc., which the user desires to see, are modelled. SOLUTION: A real robot 12 acts in a real environment 10 to be able to recognize an external environment by a sensor. A virtual environment model generating part 16 generates a three dimensional virtual environmental model 18 modelling the real environment 10. A simulator 24 simulates the sensor function which is equivalent to that of the real robot 12 in the virtual environmental model 18 to generate a virtual robot 20, and allows the virtual robot to act in synchronism with the real robot 12. A difference detecting device 26 compares sensor information on the virtual robot by the simulator 24 to sensor information on the real robot to detect a difference, and a virtual environmental model renewal part 28 renews (generates, corrects, eliminates) the virtual environmental model 18 so that the difference between the sensor information is eliminated. A monitor processing part 30 displays the virtual environmental model 18 of optional visual point and time to the user.

3. 发明专利

JPH10293566A METHOD AND DEVICE FOR PROCESSING MULTIMEDIA DATA 失效
公开(公告)号：JPH10293566A
公开(公告)日：1998-11-04
申请号：JP10212997
申请日：1997-04-18
申请人： FUJITSU LTD , GIJUTSU KENKYU KUMIAI SHINJOHO
发明人： OKADA HIROYUKI , SEKIGUCHI MINORU , WATABE NOBUO
IPC分类号： G09G5/00 , G06F17/00 , G06F19/00 , G06Q50/00 , G06Q50/04 , G06Q50/10 , G06T11/60 , G06T11/80 , G09G5/12 , G09G5/36
摘要： PROBLEM TO BE SOLVED: To make coincide the world of computer graphics with the real world in the simulation of the multimedia data that a real-time manipulation, a sound and an animation, etc., coexist. SOLUTION: A time required for plotting a display frame No is estimated, and a displayable time is calculated (2-2), and the displayable time is compared with the time to be displayed really (2-3). When the displayable time is later than the real display time (2-3-1), a display/a sound output of a relevant frame are skipped to advance to (2-5). Further, when the displayable time is equal to the real display time (2-3-2), to advance directly to the display processing (2-4) of the relevant frame. Further, when the displayable time is faster than the real display time (2-3-3), after awaiting until becoming the real display time of the relevant frame, to advance to the display processing (2-4). Then, when a final frame, the processing is ended, and when not, to return to (2-1) and the processing is performed repeatedly (2-5).

4. 发明专利

JPH10214111A DATA COLLATION METHOD, AND DATA COLLATION DEVICE AND STORAGE MEDIUM THEREFOR 失效
公开(公告)号：JPH10214111A
公开(公告)日：1998-08-11
申请号：JP32747097
申请日：1997-11-28
申请人： FUJITSU LTD
发明人： SEKIGUCHI MINORU , OKADA HIROYUKI , WATABE NOBUO , NAITO HIROHISA
IPC分类号： G05B23/02 , G05B19/4068 , G05B19/4069
摘要： PROBLEM TO BE SOLVED: To improve the accuracy of simulation data and real position data, by calculating the similarity degree of the collation object parts of the simulation data and real data, obtaining the matching part and non-matching part of both data and correcting a significant non-matching part. SOLUTION: A data collation part 5 obtains the similarity degree of the respective data elements of a time range specified in a collation object real data selection/holding part 3 and the respective data elements corresponding to the pertinent time range in the simulation data. Then, whether or not the similarity degree is ≈1 is judged, and when it is not ≈1, the similarity degree at the time of successively shifting the time range of the respective data elements to be the collation object of the simulation data to the front and back for a prescribed time width is newly calculated and the specifying processing of a maximum value in it is executed. Based on the output contents of the data collation part 5, a data correction part 6 corrects the real data position of the collation object part to simulation position data corresponding to the maximum value in the case that the maximum value is more than a reference value.

5. 发明专利

JPH1091608A DATA PROCESSOR 失效
公开(公告)号：JPH1091608A
公开(公告)日：1998-04-10
申请号：JP24086696
申请日：1996-09-11
申请人： FUJITSU LTD
发明人： OKADA HIROYUKI , WATABE NOBUO , ONDA HIROYUKI , YAMASHITA TATSUHIRO
IPC分类号： G06F15/18 , G06N3/00
摘要： PROBLEM TO BE SOLVED: To obtain a data processor which can easily recognize a deformation from an initial state by making an evaluation function set in an evaluated value calculating means have a term of a function of a coupling load when a network is in an initial state and a current initial load. SOLUTION: The inputs X and Y and output Z of a hierarchical network part 2 are coupled according to a specific fuzzy rule. A learning part 12 calculates an evaluated value by an evaluated value calculation part 13 according to the output signal of the hierachical network part 2. A network load control part 14 controls the coupling loads on respective networks of the hierarchical network part 2 according to the calculated evaluated value. An evaluation function E is set in the evaluated value calculation part 13. Here, teaching is a tutor signal, W(i, j) is a current coupling load at the coupling position i, j of a network, and ω (i, j) is the coupling load in the initial state at the coupling position i, j of the network, where ε is a constant.

6. 发明专利

JP2523151B2 失效
公开(公告)号：JP2523151B2
公开(公告)日：1996-08-07
申请号：JP789188
申请日：1988-01-18
申请人： FUJITSU LTD
发明人： OSADA SHIGEMI , ASAKAWA KAZUO , SEKIGUCHI MINORU , YOSHIZAWA HIDEKI , WATABE NOBUO , KIMOTO TAKASHI , KAWAMURA AKIRA
IPC分类号： G05B13/04 , G05B13/00 , G05B13/02 , G05B19/18 , G05B19/4155

7. 发明专利

JPH06139410A TEMPLATE MATCHING TYPE CHARACTER RECOGNITION SYSTEM USING NEURAL NETWORK 失效
公开(公告)号：JPH06139410A
公开(公告)日：1994-05-20
申请号：JP28590492
申请日：1992-10-23
申请人： FUJITSU LTD
发明人： WATABE NOBUO , OKADA HIROYUKI , KAWAMURA AKIRA
IPC分类号： G06F15/18 , G06G7/60 , G06K9/62 , G06K9/66 , G06N99/00 , G06T7/00 , G06F15/70
摘要： PURPOSE:To ensure a high recognition ratio by means of a feature value conversion neuro even though the learning of this conversion neuro is not complete by learning at a high speed the conversion neuro that converts a feature value vector into a standard one. CONSTITUTION:In a character recognition mode, a feature value vector 1 extracted from a read character pattern is converted into a standard feature value vector 3 where the feature of each category is reflected by a feature value conversion neuro 2. The template matching is carried out between the vector 3 and a template 4 and the characters are recognized. In a learning mode, a template 4 is produced from a feature vector 1' and the neuro 2 is learnt by using the data that converts the template 4 into the same one as a teacher signal. If the correct recognizing result cannot be obtained despite the learning of the neuro 2, the neuro 2 is learnt again by means of the relevant feature value vector.

8. 发明专利

JPH04275631A CONSTRUCTION SYSTEM FOR MEDIUM TYPE MEMBERSHIP FUNCTION BY NEURAL NETWORK 失效
公开(公告)号：JPH04275631A
公开(公告)日：1992-10-01
申请号：JP3598391
申请日：1991-03-01
申请人： FUJITSU LTD , FUJI FACOM CORP
发明人： WATABE NOBUO , KAWAMURA AKIRA , MASUOKA RYUSUKE , OOWADA ARIMICHI , ASAKAWA KAZUO , MATSUOKA SHIGENORI , OKADA HIROYUKI
IPC分类号： G06F9/44 , G06F15/18 , G06N3/00 , G06N7/02 , G06N99/00
摘要： PURPOSE:To approximate the upper limit of the sensitivity of a network characteristic function to the gradient of a membership function without learning with respect to the constitution system for the medium type membership function by the neural network which realizes the membership function of fuzzy control with a neuron characteristic function. CONSTITUTION:In a hierarchic network equipped with one neuron in a 1st layer which branches and outputs an input value to two, two neurons 20 and 21 in a 2nd layer which performs a threshold process by a sigmoid function for the input values from the neuron in the 1st layer, and one neuron in a 3rd layer which adds and subtracts the outputs of the neurons 20 and 21 in the 2nd layer, the weight and threshold value of the two neurons in the 2nd layer are so determined that the maximum gradient of the network characteristic function becomes equal to the gradient of the membership function medium.

9. 发明专利

JPH04275630A CONSTRUCTION SYSTEM FOR MEMBERSHIP FUNCTION BY NEURAL NETWORK 失效
公开(公告)号：JPH04275630A
公开(公告)日：1992-10-01
申请号：JP3598291
申请日：1991-03-01
申请人： FUJITSU LTD , FUJI FACOM CORP
发明人： WATABE NOBUO , KAWAMURA AKIRA , MASUOKA RYUSUKE , OOWADA ARIMICHI , ASAKAWA KAZUO , MATSUOKA SHIGENORI , OKADA HIROYUKI
IPC分类号： G06F9/44 , G06F15/18 , G06N3/00 , G06N7/02 , G06N99/00
摘要： PURPOSE:To approximate a neuron characteristic function so that a maximum error as to a membership function is minimized without learning with respect to the constitution system for the membership function by the neural network which realizes the membership function of fuzzy control with the neuron characteristic function. CONSTITUTION:The weight and threshold value of a connection are found and set so that the absolute maximum gradient of a sigmoid function as the neuron characteristic function is 1.412 time as large as the gradient of a membership function large or small.

10. 发明专利

JPH04130539A IF-THEN TYPE NETWORK CONTROL SYSTEM 失效
公开(公告)号：JPH04130539A
公开(公告)日：1992-05-01
申请号：JP25149590
申请日：1990-09-20
申请人： FUJITSU LTD , FUJI FACOM CORP
发明人： WATABE NOBUO , ASAKAWA KAZUO , OOWADA ARIMICHI , KAWAMURA AKIRA , MASUOKA RYUSUKE , MATSUOKA SHIGENORI , OKADA HIROYUKI
IPC分类号： G06F15/18 , G06F9/44 , G06N3/00 , G06N5/04 , G06N7/02 , G06N99/00
摘要： PURPOSE:To simplify and facilitate system construction and the adjustment at the time of operation by constructing all execution part networks and condition part networks with a neural network. CONSTITUTION:The system is constructed by an execution part network 10 and a condition part network 12 generating condition realization output at the time of realizing logic condition predetermined by conditions Z1 to Zm and utilizing the prewired neural network outputting the inference value of the execution part network 10 as a control value. For example, weight W of a connection 16 to final stage unit 14 is set to W=-10, and the inference value is turned to almost zero while turning the input to the final stage unit 14 of the executing part network 10 in the condition part network 12 to '-10' while the unit output is '1' against the connection 16 when the condition is not real ized in the condition part network 12. The inference value is outputted as it is while turning the input to the final stage unit 14 of the execution part net work 10 to 0 while the unit output '0' against the connection 16 when the condition is realized in the condition part network 12. Thus, the system construc tion and the adjustment at the time of operation can be simplified and facilitat ed.

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式