34 results about "Lateral inhibition" patented technology

A system including multiple devices that each have a sensor and are each configured to communicate with other devices. The system further includes a controller configured to provide command information that specifies a mode of operation of the devices. In a first mode of operation, the devices transmit communication signals and a given device modifies a strength of its communication signal from an initial strength to a final strength based on communication signals it receives from one or more other devices. And in a second mode of operation, the devices transmit communication signals and the given device dynamically adjusts a strength of its communication signal based communication signals it receives from one or more other devices and on measurements performed by the sensor in the given device.

A parallel dynamical system for computing sparse representations of data, i.e., where the data can be fully represented in terms of a small number of non-zero code elements, and for reconstructing compressively sensed images. The system is based on the principles of thresholding and local competition that solves a family of sparse approximation problems corresponding to various sparsity metrics. The system utilizes Locally Competitive Algorithms (LCAs), nodes in a population continually compete with neighboring units using (usually one-way) lateral inhibition to calculate coefficients representing an input in an over complete dictionary.

The invention relates to an image matching method based on lateral inhibition and chaos quantum particle swarm optimization, comprising the following steps of: (1) pre-processing images by using lateral inhibition principles, and extracting image edges; (2) initiating each parameter of particle swarm optimization; (3) calculating the fitness value of each particle; (4) comparing the fitness valueof each particle of the cycle, i.e. the kth cycle, and the fitness value of each particle obtained from the former step, wherein the maximal fitness value is the global optimization gbest; simultaneously comparing the fitness value of each particle obtained from the former step respectively with the fitness value of the particle obtained in the (k-1)th cycle, wherein the greater fitness value is the optimization Pbest of the particle self of the kth cycle; (5) carrying out particle swarm optimization according to the fitness values of the particles; (6) skipping to the step 3 when the currentcycle time plus 1, repeating the operation till the cycle time is greater than the maximum cycle time N; and (7) ending the operation and outputting the optimized allocation position and the optimized fitness value.

The invention provides an improved top speed learning model and a method for classifying modes of the top speed learning model, discloses an efficient neural network rapid learning method and a limited top speed learning machine and belongs to the field of mode identification and machine learning. The improved top speed learning model comprises (1) a concept of a limited parameter space, (2) a hypersphere limited condition in the limited parameter concept and (3) output weight learning. According to the concept of the limited parameter space, a lateral inhibition mechanism based on sample prior information is adopted and the content is particularly shown on the aspect of generation of a connection weight from an input layer to a hidden layer; according to the hypersphere limited condition in the limited parameter concept, selection of the connection weight from the input layer to the hidden layer is limited to a hypersphere; according to output weight learning, after a weight of the hidden layer is selected from the limited parameter space, learning is trained through a top speed learning machine model based on least squares, and finally an output weight of the model is obtained. According to the method, the classification and identification effects of the model can be greatly improved, and the training speed can be greatly increased.

The invention discloses a moving target position detection system and method. The method includes: calibrating the visual axis of the sub-sensor unit in the multi-channel data synchronous acquisition device by Gaussian fitting method; according to the multi-channel data synchronous acquisition device having calibrated the sub-sensor unit visual axis, a side-inhibition neural network algorithm based on signal weighting , detecting the moving angle and moving distance of the target to be measured; determining the position of the target to be measured according to the moving angle and the moving distance. The present invention uses the quasi-Gaussian state of the target signal in the sensor array to propose a fast calibration method for the visual axis of the sensor array; based on the lateral inhibition characteristics of the optic nerve of the compound eye, combined with a weighting algorithm, the super-resolution of the motion parameters of the environmental target is realized under the high time resolution of the visual sensor High-rate angle positioning and ranging, so that it can quickly and accurately detect the position of moving targets.

The present invention proposes a facial age recognition system and method based on a pulse neural network, and a system for performing facial age recognition based on an unsupervised facial feature vector, a computer program, and a computer-readable storage medium for performing the above method, which can be used dynamically Good results have been achieved on the dataset. Compared with traditional methods, the technical scheme of the present invention uses spiking neurons to build a spiking neural network model, and uses the STDP rule (pulse time-dependent synaptic plasticity rule) to train the spiking neural network, and uses a lateral inhibition method , obtained an unsupervised self-organizing map neural network, which can continuously adapt to changes in actual data. Therefore, the problem that the existing technology is only applicable to static data sets is solved.

The invention relates to a lateral inhabitation coefficient determining method based on unilateral attenuation exponential and belongs to the field of image enhancement processing. A lateral inhibition coefficient distribution function is based on the unilateral attenuation exponential, the size of the radius N of a lateral inhibition field range is used for determining a coefficient sigma, the lateral inhibition coefficient is distributed and concentrated in a small lateral inhibition field range through controlling the lateral inhibition coefficient attenuation speed rate to reduce calculation amount of a lateral inhibition network, and a coefficient A is determined based on the optimal lateral inhibition coefficient and the sigma to effectively enhance image visual effects. The method is simple in implementation process, small in calculation amount and obvious in image enhancement effect, and has application and popularization value.

The invention relates to a contour detection method on the basis of a primary visual pathway calculation model. A primary visual pathway computation model is constructed, and contour detection is realized by simulating the transmission and processing process of visual information flows. A classical receptive field direction selection model combining multi-scale features is provided, and a multi-scale feature fusion strategy is utilized to simulate the primary contour response of the retinal ganglion. In a visual pathway from the retinal ganglion to the lateral geniculate nucleus, a space-timecoding mechanism is adopted for reducing redundant features in the primary contour response. A computation model with non-classical receptive field isotropic suppression properties is provided, and bymeans of a synergistic effect of non-subsampled contourlet transform and Gabor transform, a processing effect of lateral suppression characteristics of a non-classical receptive field on texture background information is simulated; a feedforward mechanism of a primary visual pathway to the primary visual cortex is simulated, visual characteristics of multiple visual pathways are fused, and the contour response is finally obtained. According to the contour detection method, a main body contour can be effectively highlighted, and a texture background can be suppressed.

The invention discloses a probabilistic neuron circuit, a topological structure of a probabilistic neural network and an application thereof. The probabilistic neuron circuit comprises an integral capacitor, a non-fixed threshold volatile device and a load resistor; One end of the fixed-threshold volatile device, and the other end of the volatile device is connected to one end of the load resistor. The network topology includes multiple input neuron circuits, multiple probabilistic neuron circuits, and lateral inhibitory neuron circuits; each probabilistic neuron circuit is used to perform a neural network based on its non-fixed excitation threshold and the electrical signals emitted by each input neuron circuit. Random excitation; the inhibitory neuron circuit is used to inhibit the excitation of other subsequent probabilistic neuron circuits when receiving the signals excited by the first n probability neuron circuits. The invention introduces a non-fixed threshold volatile device into the neuron circuit, which greatly expands the application of the neuron circuit, especially can be used to solve non-deterministic problems, and has reliable solution results.

The invention discloses an image edge classification detection method based on visual path orientation sensitivity. The present invention utilizes the role of neuron synaptic connections in the centripetal distribution of the optimal orientation of the receptive field, constructs a first-level neuron network sensitive to multiple specific directions, uses image pixels as network input, and records neurons in a certain time window The pulse firing sequence within is calculated, and the discharge frequency is calculated as the network output; the network output in multiple directions is fused and mapped to the gray level to form an edge-sensitive image; for the edge-sensitive image, the lateral inhibition range and inhibition amount in the receptive field are determined , form the second-level neuron network, and output the image after lateral suppression; finally, after threshold processing, the edge detection result is obtained. The invention considers important visual mechanisms such as directional receptive fields and lateral inhibition, simulates the hierarchical processing effect of different hierarchical structures on the visual path in contour perception, and can effectively improve the edge detection performance of low-contrast images.

The present invention relates to the technical field of computer vision, specifically provides a visual attention detection method and device based on lateral inhibition, and aims to solve how to improve the effectiveness and practicability of computer vision attention mechanism. For this purpose, the visual attention detection method among the present invention comprises the steps of: obtaining the image category of the target image; obtaining the feedback information received by each activation function neuron; judging whether to close the corresponding activation function neuron according to the feedback information; After closing the corresponding activation function neuron according to the judgment result, the response value of the activation function neuron is normalized to obtain the selective attention result map; the selective attention result map is normalized to obtain the salient target detection Result graph. The technical scheme of the present invention is based on the convolutional neural network, and can utilize the feedforward and feedback information of the convolutional neural network to realize the information detection of the target object. Meanwhile, the device in the present invention can execute and realize the above steps.