公开(公告)号：EP0768610A1

公开(公告)日：1997-04-16

申请号：EP95830433.9

申请日：1995-10-13

Applicant: SGS-THOMSON MICROELECTRONICS S.r.l.

Inventor： Fabbrizio, Vito ,  Colli, Gianluca ,  Kramer, Alan

IPC: G06F15/80 ,  G06G7/60

CPC classification number: G06N3/063 ,  G06N3/0635

Abstract: A neural network (1) including a number of synaptic weighting elements (15, 17), and a neuron stage (5); each of the synaptic weighting elements (15, 17) having a respective synaptic input connection (11, 13) supplied with a respective input signal (x 1 , ..., x n ); and the neuron stage (5) having inputs (36, 37) connected to the synaptic weighting elements, and being connected to an output (39) of the neural network (1) supplying a digital output signal (O). The synaptic weighting elements (15, 17) are formed by memory cells programmable to different threshold voltage levels, so that each presents a respective programmable conductance; and the neuron stage (5) provides for measuring conductance (33-35, 43-45) on the basis of the current through the memory cells, and for generating a binary output signal on the basis of the total conductance of the synaptic elements.

Abstract translation: 包括许多突触加权元素（15,17）和神经元阶段（5）的神经网络（1）; 每个突触加权元件（15,17）具有各自的输入信号（x1，...，xn）的突触输入连接（11,13）; 和具有与突触加权元件连接的输入端（36,37）的神经元级（5），并连接到提供数字输出信号（O）的神经网络（1）的输出端（39）。 突触加权元件（15,17）由可编程为不同阈值电压电平的存储器单元形成，使得每个呈现相应的可编程电导; 并且神经元阶段（5）基于通过存储器单元的电流提供测量电导（33-35,43-45），并且基于突触元件的总电导产生二进制输出信号。

公开(公告)号：EP0837399A1

公开(公告)日：1998-04-22

申请号：EP96830525.0

申请日：1996-10-15

Applicant: SGS-THOMSON MICROELECTRONICS s.r.l.

Inventor： Fabbrizio, Vito ,  Kramer, Alan

IPC: G06F15/80

CPC classification number: G06N3/063

Abstract: An electronic device (100) for performing convolution operations comprises shift registers (106-120) for receiving binary input values (122-129) representative of an original matrix, synapses (142) for storing weights correlated with a mask matrix, and neurones (154, 156) for outputting (166, 168) a binary result dependent on the sum of the binary values weighted by the synapses (142), each synapse (142) having a conductance correlated with the weight stored and dependent upon the binary input value and each neurone (154, 156) generating the binary result in dependence on the total conductance of the corresponding synapses (142).

Abstract translation: 用于执行卷积运算的电子设备（100）包括用于接收代表原始矩阵的二进制输入值（122-129）的移位寄存器（106-120），用于存储与掩码矩阵相关的权重的突触（142）和神经元 用于输出（166,168）二进制结果，二进制结果取决于由突触（142）加权的二进制值之和，每个突触（142）具有与存储的权重相关的电导，并且取决于二进制输入值 并且每个神经元（154,156）根据相应突触的总电导（142）产生二进制结果。

公开(公告)号：EP0837378A2

公开(公告)日：1998-04-22

申请号：EP97203126.4

申请日：1997-10-08

Applicant: SGS-THOMSON MICROELECTRONICS s.r.l.

Inventor： Olivieri, Massimiliano ,  Fabbrizio, Vito ,  Guerrieri, Roberto ,  Kramer, Alan

IPC: G05B1/03

CPC classification number: G05D1/0246 ,  G05D2201/0213

Abstract: Method (10), in a system for aiding the guidance of a vehicle, for identifying marking stripes of road lanes comprising the phases of subjecting a road image to a convolution operation (14) with a mask matrix so as to identify discontinuities present in the image, comparing the result with a threshold value (16) and determining (18) a representation of the marking stripes, in which the mask matrix is set in such a way as to eliminate at least partially the discontinuities which do not correspond to the marking stripes.

Abstract translation: 方法（10）在用于辅助车辆的引导的系统中，用于识别道路车道的标记条纹，其包括用道路图像进行带有掩模矩阵的卷积操作（14）的阶段，以便识别存在于 图像，将结果与阈值（16）进行比较并确定（18）标记条纹的表示，其中掩模矩阵被设置为至少部分地消除不对应于标记的不连续性 条纹。

公开(公告)号：EP0837378A3

公开(公告)日：1998-05-06

申请号：EP97203126.4

申请日：1997-10-08

Applicant: SGS-THOMSON MICROELECTRONICS s.r.l.

Inventor： Olivieri, Massimiliano ,  Fabbrizio, Vito ,  Guerrieri, Roberto ,  Kramer, Alan

IPC: G05B1/03

CPC classification number: G05D1/0246 ,  G05D2201/0213

Abstract: Method (10), in a system for aiding the guidance of a vehicle, for identifying marking stripes of road lanes comprising the phases of subjecting a road image to a convolution operation (14) with a mask matrix so as to identify discontinuities present in the image, comparing the result with a threshold value (16) and determining (18) a representation of the marking stripes, in which the mask matrix is set in such a way as to eliminate at least partially the discontinuities which do not correspond to the marking stripes.

Abstract translation: 方法（10），在用于辅助车辆引导的系统中，用于识别道路车道的标记条纹，包括使道路图像经受与掩模矩阵的卷积运算（14）的阶段，以识别存在于 图像，将结果与阈值（16）进行比较并确定（18）标记条纹的表示，其中设置掩模矩阵以至少部分消除不对应于标记的不连续性 条纹。