Abstract: A neural network apparatus (20) includes a storage unit (24) storing a neural network model, and an arithmetic unit (22) inputting input information into an input layer of the neural network and outputting an output layer. A weight matrix (W) of an FC layer of the neural network model is constituted by a product of a weight basis matrix (Mw) of integers and a weight coefficient matrix (Cw) of real numbers. In the FC layer, the arithmetic unit (22) uses an output vector from a previous layer as an input vector (x) to decompose the input vector (x) into a product of a binary input basis matrix (Mx) and an input coefficient vector (cx) of real numbers and an input bias (bx) and derives a product of the input vector (x) and a weight matrix (W).

Abstract: The image restoration device has an initialization block that initializes the luminance value of each pixel coordinate to an intermediate value in the luminance array list that stores one of a pair of polarity values and intermediate values as the luminance value for each pixel coordinate. The image restoration device also has an update block that updates the initialized luminance array list according to the pixel coordinates and polarity values for each event, and an output block that outputs the luminance array list updated by the update block over the shooting period as a binary image. By the update performed in the update block, the luminance values of the firing coordinates where the event fired in the luminance array list are overwritten by the polarity values of the event. In addition, the update preserves the luminance values of the non-firing coordinates in the luminance array list, excluding the firing coordinates.

Abstract: According to one embodiment of the present disclosure, provided is a learning system that updates a parameter for a neural network, the learning system including: a plurality of differential value calculators; and a parameter update module.

Abstract: Disclosed is a technique of determining a driver's tension level (tension state degree) in vehicle driving in detail with a simple configuration. According to the technique, a nonlinear analyzing unit 110 of a driver's tension level determining apparatus 100 determining the tension level in driving of a driver acquires the driving operation amounts relating to driving operations of a driver (the operation amounts relating to operations of an accelerator pedal, a brake pedal, a handle, and the like), and then calculates the Lyapunov exponents about the driving operation amounts by performing nonlinear analysis processing. A frequency spectrum analyzing unit 120 calculates the power spectral density of time series data of the Lyapunov exponents, and then calculates an integrated value of a predetermined low frequency band in the calculated power spectral density.

Abstract: A recognition system includes: a sensor processing unit (SPU) that performs sensing to output a sensor value; a task-specific unit (TSU) including an object detection part that performs an object detection task based on the sensor value and a semantic segmentation part that performs a semantic segmentation task based on the sensor value; and a generic-feature extraction part (GEU) including a generic neural network disposed between the sensor processing unit and the task-specific unit, the generic neural network being configured to receive the sensor value as an input to extract a generic feature to be input in common into the object detection part and the semantic segmentation part.

Abstract: A technique of estimating a surrounding environment with higher accuracy from observation data obtained by using sensing equipment is disclosed. According to the technique, a certain pixel in an observation data space which includes the observation data obtained by a sensor 50 is sampled. With regard to the sampled pixel, a grid-unspecified logarithm-likelihood ratio indicating whether a received signal is a signal reflected from a detected object is calculated from an intensity value of the received signal, and a distribution function of an angle indicating the degree of dispersion centering on an azimuth angle is calculated. Then, the product of the grid-unspecified logarithm-likelihood ratio and the distribution function is calculated and thereby a calculation for updating an occupation probability of the object is performed across a plurality of grids in an occupancy grid map by using a value obtained by dispersing the likelihood using the distribution function.

Abstract: A technique for stable and fast computation of a variance representing a confidence interval for an estimation result in an estimation apparatus using a neural network including an integrated layer that combines a dropout layer for dropping out part of input data and an FC layer for computing a weight is provided. When input data having a multivariate distribution is supplied to the integrated layer, a data analysis unit 30 determines, based on a numerical distribution of terms formed by respective products of each vector element of the input data and the weight, a data type of each vector element of output data from the integrated layer. An estimated confidence interval computation unit 20 applies an approximate computation method associated with the data type, to analytically compute a variance of each vector element of the output data from the integrated layer based on the input data to the integrated layer.

Abstract: The moving object recognition system includes: a camera that is installed in a vehicle and captures continuous single-view images; a moving object detecting unit that detects a moving object from the images captured by the camera; a relative approach angle estimating unit that estimates the relative approach angle of the moving object detected by the moving object detecting unit with respect to the camera; a collision risk calculating unit that calculates the risk of the moving object colliding with the vehicle, based on the relationship between the relative approach angle and the moving object direction from the camera toward the moving object; and a reporting unit that reports a danger to the driver of the vehicle in accordance with the risk calculated by the collision risk calculating unit.

Abstract: Disclosed is a technique of determining a driver's tension level (tension state degree) in vehicle driving in detail with a simple configuration. According to the technique, a nonlinear analyzing unit 110 of a driver's tension level determining apparatus 100 determining the tension level in driving of a driver acquires the driving operation amounts relating to driving operations of a driver (the operation amounts relating to operations of an accelerator pedal, a brake pedal, a handle, and the like), and then calculates the Lyapunov exponents about the driving operation amounts by performing nonlinear analysis processing. A frequency spectrum analyzing unit 120 calculates the power spectral density of time series data of the Lyapunov exponents, and then calculates an integrated value of a predetermined low frequency band in the calculated power spectral density.

Abstract: A technique of estimating a surrounding environment with higher accuracy from observation data obtained by using sensing equipment is disclosed. According to the technique, a certain pixel in an observation data space which includes the observation data obtained by a sensor 50 is sampled. With regard to the sampled pixel, a grid-unspecified logarithm-likelihood ratio indicating whether a received signal is a signal reflected from a detected object is calculated from an intensity value of the received signal, and a distribution function of an angle indicating the degree of dispersion centering on an azimuth angle is calculated. Then, the product of the grid-unspecified logarithm-likelihood ratio and the distribution function is calculated and thereby a calculation for updating an occupation probability of the object is performed across a plurality of grids in an occupancy grid map by using a value obtained by dispersing the likelihood using the distribution function.

Abstract: Provided is a traffic congestion prediction method which is able to perform a prediction process using floating information with higher accuracy. The traffic congestion prediction method includes: a step of receiving information by a prediction device; a step of predicting a route of each floating car based on the current position information and destination information received; a step of calculating, for the each floating car, a first passing time group which is a set of respective passing times at a plurality of predetermined spots on the route predicted; a step of calculating the number of existing floating cars per link based on the first passing time group, if any of a plurality of floating cars exists on the link at a predetermined time; and a step of calculating a second passing time group by use of the number of existing floating cars and a predetermined calculation technique.

Abstract: A traffic data prediction device includes an original link traffic data storage unit (103) for storing traffic data per original link as a predetermined road link, an extended link generation unit (104) for generating an extended link from the original links, and an extended link traffic data prediction unit (108) for predicting traffic data per extended link generated in the extended link generation unit (104) by use of traffic data per original link. The extended link generation unit (104) decides the original links for generating the extended link based on data indicating a predictive accuracy of traffic data in a combined link combining the selected original links, and generates the extended link made of the decided original links as elements.

Abstract: A local feature amount calculating device includes a local feature amount calculating unit that acquires a local feature amount of a feature point for each pixel by giving a vote of a strength of the direction of a pixel, which is calculated by a direction strength calculating unit, to a vote cell to which a pixel of interest belongs out of vote cells of a rotated vote cell pattern that is acquired by rotating a reference vote cell pattern by an angle corresponding to a main axis direction along a corrected edge gradient direction that is acquired by correcting an edge gradient direction of the pixel that is calculated by a direction strength calculating unit in accordance with the main axis direction.

Abstract: An object of this invention is to improve the accuracy of estimating a destination in a destination estimating apparatus. A destination estimating apparatus 100 includes: a learning data storing unit 9b that stores a history of a location specified as a destination in the past; a destination estimating unit 83 that estimates a destination from among a plurality of destination candidates including a location stored in the learning data storing unit 9b; and a candidate excluding unit 84 that, based on the history stored in the learning data storing unit 9b, excludes a destination candidate for which it is determined that a certainty factor of being a destination is lower than a predetermined threshold value from destination candidates that are estimated as being a destination by the destination estimating unit 83.

Abstract: A traffic data prediction device includes an original link traffic data storage unit (103) for storing traffic data per original link as a predetermined road link, an extended link generation unit (104) for generating an extended link from the original links, and an extended link traffic data prediction unit (108) for predicting traffic data per extended link generated in the extended link generation unit (104) by use of traffic data per original link. The extended link generation unit (104) decides the original links for generating the extended link based on data indicating a predictive accuracy of traffic data in a combined link combining the selected original links, and generates the extended link made of the decided original links as elements.

Abstract: The moving object recognition system includes: a camera that is installed in a vehicle and captures continuous single-view images; a moving object detecting unit that detects a moving object from the images captured by the camera; a relative approach angle estimating unit that estimates the relative approach angle of the moving object detected by the moving object detecting unit with respect to the camera; a collision risk calculating unit that calculates the risk of the moving object colliding with the vehicle, based on the relationship between the relative approach angle and the moving object direction from the camera toward the moving object; and a reporting unit that reports a danger to the driver of the vehicle in accordance with the risk calculated by the collision risk calculating unit.

Abstract: A bit code converter transforms a learning feature vector using a transformation matrix updated by a transformation matrix update unit, and converts the transformed learning feature vector into a bit code. When the transformation matrix update unit substitutes a substitution candidate for an element of the transformation matrix, a cost function calculator fixes the substitution candidate that minimizes a cost function as the element. The transformation matrix update unit selects the element while sequentially changing the elements, and the cost function calculator fixes the selected element every time the transformation matrix update unit selects the element, thereby finally fixing the optimum transformation matrix.

Abstract: Provided is a traffic congestion prediction method which is able to perform a prediction process using floating information with higher accuracy. The traffic congestion prediction method includes: a step of receiving information by a prediction device; a step of predicting a route of each floating car based on the current position information and destination information received; a step of calculating, for the each floating car, a first passing time group which is a set of respective passing times at a plurality of predetermined spots on the route predicted; a step of calculating the number of existing floating cars per link based on the first passing time group, if any of a plurality of floating cars exists on the link at a predetermined time; and a step of calculating a second passing time group by use of the number of existing floating cars and a predetermined calculation technique.

Abstract: A local feature amount calculating device includes a local feature amount calculating unit that acquires a local feature amount of a feature point for each pixel by giving a vote of a strength of the direction of a pixel, which is calculated by a direction strength calculating unit, to a vote cell to which a pixel of interest belongs out of vote cells of a rotated vote cell pattern that is acquired by rotating a reference vote cell pattern by an angle corresponding to a main axis direction along a corrected edge gradient direction that is acquired by correcting an edge gradient direction of the pixel that is calculated by a direction strength calculating unit in accordance with the main axis direction.

Abstract: A technology that makes a program line-up related to contents distributed to a user side, depending on various conditions, preferences, and communication environments of the user side that views and listens to the content is disclosed. According to the technology, an on-board device (content receiving and reproducing device) 1 receives, from a service server 5, potential content list information of contents that can be distributed from a content server 7. The on-board device 1 sorts appropriate content from among the contents in the potential content list information and decide on a reproducing order of the contents, based on conditions of the user side, such as user preferences and vehicle conditions, conditions related to an environment on the user side, such as the communication environment, and conditions related to the contents, such as the genre of the content. The on-board device 1 creates program table information (timetable) and transmits the program table information to the content server.