Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

An image recognition method based on hierarchical feature extraction and multi-layer impulse neural network

A technology of spiking neural network and image recognition, applied in the field of spiking neural network, it can solve the problems of difficult feedback and discontinuous pulse delivery.

Active Publication Date: 2018-12-28
HANGZHOU DIANZI UNIV
View PDF6 Cites 14 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Using probabilistic methods to solve the problem of discontinuous pulse firing in multi-layer spiking neural networks, which leads to difficult feedback

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • An image recognition method based on hierarchical feature extraction and multi-layer impulse neural network
  • An image recognition method based on hierarchical feature extraction and multi-layer impulse neural network
  • An image recognition method based on hierarchical feature extraction and multi-layer impulse neural network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0052] The present invention will be further described below in conjunction with the accompanying drawings.

[0053] Figure 1 to Figure 4 Each stage of the entire image recognition process is shown separately. The process can be divided into 3 steps, and the specific content is as follows:

[0054] Step 1 Hierarchical Feature Extraction

[0055] in figure 2 The overall process of hierarchical feature extraction is described. In this process, a four-layer model is adopted, which are S1 layer, C1 layer, S2 layer and C2 layer. The parameter values ​​involved here are mainly for the MNIST data set. The specific operations of each layer are as follows:

[0056] 1.1S1 layer: extraction of edge information by Gabor filter

[0057] The cells in the primary visual cortex area are strongly sensitive to edge information, and the frequency and direction expression of the Gabor filter are considered to be similar to the human visual system, so in this step a two-dimensional Gabor fil...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses an image recognition method based on hierarchical feature extraction and multi-layer impulse neural network. According to the processing mode of the visual information by the visual cortex, on the basis of a HMAX model, a sparse feature and feature autonomous learning method is introduced to make the hierarchical feature extraction result retain valid information reasonably, and a multi-layer impulse neural network model based on STDP and back propagation algorithm is used to realize the training and recognition of extracted data. Moreover, phase coding is used as a bridge between hierarchical feature extraction and multi-layer impulse neural network, which effectively converts pixel information into time information and improves the recognition accuracy. The imagerecognition method of the invention not only satisfies the biological characteristics but also has good classification performance. In the process of hierarchical feature extraction, the combination of manual features and autonomous learning features can better meet different needs; at the same time, using multi-layer impulse neural network to identify and classify the complex data can be effectively processed.

Description

technical field [0001] The invention relates to the field of impulse neural networks, in particular to an image recognition method based on layered feature extraction and multi-layer impulse neural networks. Background technique [0002] At present, a series of computational models have been used to simulate the input-output relationship of visual information in the visual cortical area. These models focus on the representation of information, motion state, color texture, etc. or focus on specific functions, such as object recognition, boundary detection, action recognition, etc. While these models explain the underlying computational mechanisms of vision, they lack biologically grounded explanations. In order to simulate the high-efficiency and low-power characteristics of the visual information processing of the brain's visual cortex, the biological neural network is applied to the computer vision computing model. Based on the biological principles of the visual system, b...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
IPC IPC(8): G06K9/62G06N3/04G06N3/06G06N3/08
CPCG06N3/061G06N3/08G06N3/045G06F18/2136G06F18/241
Inventor 徐小良卢文思方启明
Owner HANGZHOU DIANZI UNIV
Features
  • Generate Ideas
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More