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

JP2011065545A Device and method for signal identification 有权
标题翻译：用于信号识别的装置和方法
公开(公告)号：JP2011065545A
公开(公告)日：2011-03-31
申请号：JP2009217334
申请日：2009-09-18
申请人： Panasonic Electric Works Co Ltd , パナソニック電工株式会社
发明人： HASHIMOTO YOSHIHITO
IPC分类号： G06N3/00
摘要： PROBLEM TO BE SOLVED: To reduce an erroneous decision in which normal conditions are determined as abnormal conditions while preventing increase of the erroneous decision in which the abnormal conditions are determined as the normal conditions without relaxing criteria for conditions of an inspection object, and also to largely reduce the inspection time. SOLUTION: The signal identification device forms a clustering map by causing a neural network 1 to learn by use of a plurality of learning data. In a signal input unit 2, a vibration sensor 21 detects and converts a vibration of the inspection object A to an inspection signal, and a microphone 22 detects and converts a sound of the inspection object into an inspection signal. A feature quantity extraction unit 3 acquires the inspection signals from the vibration sensor 21 and the microphone 22, respectively, and extracts a feature quantity showing a relationship of a plurality of acquired inspection signals as inspection data. In an arithmetic unit 5, a cluster determination unit 52 inputs the inspection data to the learnt neural network 1 to determine the conditions of the inspection object A from the position of the inspection data on the clustering map. COPYRIGHT: (C)2011,JPO&INPIT
摘要翻译：要解决的问题为了减少将正常条件确定为异常状况的错误决定，同时防止异常状况被确定为正常条件的错误判定的增加而不放松对检查对象的条件的标准，并大大减少检验时间。解决方案：信号识别装置通过使神经网络1通过使用多个学习数据来学习来形成聚类图。在信号输入单元2中，振动传感器21检测检测对象A的振动并将其转换为检查信号，麦克风22检测并将检查对象的声音转换为检查信号。特征量提取单元3分别从振动传感器21和麦克风22获取检测信号，并提取表示多个获取的检查信号的关系的特征量作为检查数据。在算术单元5中，群集确定单元52将检查数据输入到学习神经网络1，以从聚类图上的检查数据的位置确定检查对象A的状况。版权所有（C）2011，JPO＆INPIT

2. 发明专利

JP2011090627A Signal identification method and signal identification apparatus 有权
标题翻译：信号识别方法和信号识别装置
公开(公告)号：JP2011090627A
公开(公告)日：2011-05-06
申请号：JP2009245537
申请日：2009-10-26
申请人： Panasonic Electric Works Co Ltd , パナソニック電工株式会社
发明人： HASHIMOTO YOSHIHITO
IPC分类号： G06N3/00 , G06N3/08
摘要： PROBLEM TO BE SOLVED: To automatically select a combination of extraction areas being ares to be inspected, and to improve the accuracy of inspection.
SOLUTION: During a learning phase, an extraction unit 3 extracts a feature value of each learning signal for each extraction range and a selection unit 4 selects an area to be inspected. A setting unit 5 creates a clustering map for each extraction range of the area to be inspected by providing learning data which are a product of feature values with a degree of confidence to a neural network. Then, the setting unit 5 calculates a Euclidean distance between a learnt weight vector and the learning data to select a maximal value and sets a Gaussian function. Then, the setting unit 5 sets a threshold for inspection by calculating a total of Gaussian function values for the learning data. During an inspection phase, an inspection unit 7 selects the feature value of the extraction range of the area to be inspected. Then, the inspection unit 7 calculates a total of the Gaussian function values using inspection data which are a product of the feature values with the degree of confidence and determines that the object to be inspected A is in a normal state if the total of the Gaussian function values is not smaller than the threshold for inspection.
COPYRIGHT: (C)2011,JPO&INPIT
摘要翻译：要解决的问题：自动选择要检查的区域的提取区域的组合，并提高检查的准确性。解决方案：在学习阶段，提取单元3针对每个提取范围提取每个学习信号的特征值，并且选择单元4选择要检查的区域。设置单元5通过向神经网络提供作为具有置信度的特征值的乘积的学习数据，来创建要检查的区域的每个提取范围的聚类映射。然后，设置单元5计算学习权重向量和学习数据之间的欧几里德距离，以选择最大值并设置高斯函数。然后，设定单元5通过计算学习数据的高斯函数值的总和来设定检查阈值。在检查阶段期间，检查单元7选择要检查的区域的提取范围的特征值。然后，检查单元7使用作为特征值与置信度的乘积的检查数据来计算高斯函数值的总和，并且如果高斯函数的总和确定被检查对象A处于正常状态功能值不小于检查阈值。版权所有（C）2011，JPO＆INPIT

3. 发明专利

JP2009146149A Signal identification method and signal identification device 审中-公开
标题翻译：信号识别方法和信号识别装置
公开(公告)号：JP2009146149A
公开(公告)日：2009-07-02
申请号：JP2007322474
申请日：2007-12-13
申请人： Panasonic Electric Works Co Ltd , パナソニック電工株式会社
发明人： HASHIMOTO YOSHIHITO
IPC分类号： G06N3/00 , G06N3/08
摘要： PROBLEM TO BE SOLVED: To shorten a processing time necessary for the setting of a data set for learning without deteriorating precision.
SOLUTION: An accuracy arithmetic part 60 of a signal processing part 6 calculates the total sum of complicatedness for each selection data set created by a featured value extraction part 2, and a data creation part 61 selects two selection data sets from the selection data sets for which the total sum of complicatedness has been calculated by using genetic algorithm, and generates a new selection data set by carrying out crossing processing to the selected selection data set, or performing mutation processing by a preset probability. Afterward, the accuracy arithmetic part 60 calculates the total sum of complicatedness for each new selection data. When all generations are ended, a learning data set setting part 62 rearranges the selection data sets in the order of the small total sum of complicatedness, that is, in the order of high adaptability, and sets the selection data set whose adaptability is the highest as a data set for learning.
COPYRIGHT: (C)2009,JPO&INPIT
摘要翻译：要解决的问题：缩短设置用于学习的数据集所需的处理时间，而不会降低精度。解决方案：信号处理部分6的精度算术部分60计算由特征值提取部分2创建的每个选择数据集的复杂度的总和，数据创建部分61从选择中选择两个选择数据集通过使用遗传算法计算出复杂度的总和的数据集，并且通过对所选择的选择数据集进行交叉处理，或者以预设的概率进行突变处理，生成新的选择数据集。之后，精度运算部60计算各新的选择数据的复杂度的总和。当所有世代结束时，学习数据集设定部分62以小的复杂度的总和，即适应性高的顺序重新排列选择数据集，并且设置适应性最高的选择数据集作为学习的数据集。版权所有（C）2009，JPO＆INPIT

4. 发明专利

JP2010231455A Method and device for signal identification 审中-公开
标题翻译：用于信号识别的方法和装置
公开(公告)号：JP2010231455A
公开(公告)日：2010-10-14
申请号：JP2009077688
申请日：2009-03-26
申请人： Panasonic Electric Works Co Ltd , パナソニック電工株式会社
发明人： HASHIMOTO YOSHIHITO
IPC分类号： G06N5/04 , G01M99/00 , G06N3/00
CPC分类号： G06K9/6284 , G06K9/00536 , G10L25/30 , G10L25/78
摘要： PROBLEM TO BE SOLVED: To automatically achieve learning without any special knowledge, and to shorten an inspection time.
SOLUTION: In a signal identification method, when a signal for learning including a normal sample and an abnormal sample is input in learning, a feature value extraction unit 2 performs short-time Fourier transformation to the signal for learning, and extracts data for learning. An identifier creation unit 6 creates an identifier to minimize an error determination rate calculated by a calculation unit 5 by using the determination result of a learning time decision unit 4 for each of the combinations of time and frequency. An identifier selection unit 7 selects an identifier whose error determination rate is minimum from identifiers created for each of the combinations of the time and frequency, and calculates reliability. A weighting instruction unit 31 instructs a weighting setting change unit 30 to change the weighting of data for learning according to the determination result of the selected identifier. In inspection, an inspection time decision unit 8 determines whether a test object is put in a normal state by using a plurality of identifies selected in learning.
COPYRIGHT: (C)2011,JPO&INPIT
摘要翻译：要解决的问题：无需任何专门知识自动实现学习，缩短检验时间。解决方案：在信号识别方法中，当在学习中输入包括正常样本和异常样本的用于学习的信号时，特征值提取单元2对学习信号执行短时傅里叶变换，并且提取数据学习。标识符创建单元6通过使用用于每个时间和频率的组合的学习时间决定单元4的确定结果来创建由计算单元5计算出的错误确定率最小化的标识符。标识符选择单元7从为时间和频率的组合中的每一个创建的标识符中选择错误确定率最小的标识符，并且计算可靠性。加权指令单元31指示加权设置改变单元30根据所选标识符的确定结果改变用于学习的数据的加权。在检查中，检查时间决定单元8通过使用在学习中选择的多个标识来确定测试对象是否处于正常状态。版权所有（C）2011，JPO＆INPIT

5. 发明专利

JP2010008159A Visual inspection processing method 审中-公开
标题翻译：视觉检测处理方法
公开(公告)号：JP2010008159A
公开(公告)日：2010-01-14
申请号：JP2008166272
申请日：2008-06-25
申请人： Panasonic Electric Works Co Ltd , パナソニック電工株式会社
发明人： HASHIMOTO YOSHIHITO
IPC分类号： G01N21/88 , G06T1/00 , G06T1/40 , G06T7/00
摘要： PROBLEM TO BE SOLVED: To quickly and precisely make a visual inspection. SOLUTION: A color extraction section 5 creates learning data with each coordinates value of RGB as the amount of feature for each pixel for RGB color images where an inspection target in a normal state has been captured. A classification section 2 inputs respective learning data into a competitive learning type neural network 1 for learning to create a clustering map. After learning, the classification section 2 obtains Euclidean distance at each neuron on the clustering map for respective learning data inputted to the competitive learning type neural network 1 again, and creates a list of Euclidean distances for each neuron. After that, the classification section 2 sets a Gaussian function where a weight vector is defined as an average vector with the maximum value of the list as variance for each neuron. Then, the classification section 2 obtains the total Gaussian function value of all neurons for respective learning data and sets the minimum value of the total Gaussian function value relating to all learning data to a lower limit threshold. COPYRIGHT: (C)2010,JPO&INPIT
摘要翻译：要解决的问题：快速准确地进行目视检查。解决方案：颜色提取部分5创建具有RGB的每个坐标值的学习数据作为捕获正常状态的检查对象的RGB彩色图像的每个像素的特征量。分类部分2将相应的学习数据输入到用于学习创建聚类映射的竞争性学习型神经网络1中。在学习之后，分类部分2再次在输入到竞争性学习型神经网络1的各个学习数据的聚类图上的每个神经元处获得欧几里得距离，并为每个神经元创建欧几里得距离列表。之后，分类部分2设置高斯函数，其中权重向量被定义为具有列表的最大值作为每个神经元的方差的平均向量。然后，分类部分2获得各个学习数据的所有神经元的总高斯函数值，并将与所有学习数据相关的总高斯函数值的最小值设置为下限阈值。版权所有（C）2010，JPO＆INPIT

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