A Scale and Translation Invariant Approach for Early Classification of Spatio-Temporal Patterns Using Spiking Neural Networks

被引:0
作者
Banafsheh Rekabdar
Monica Nicolescu
Mircea Nicolescu
Mohammad Taghi Saffar
Richard Kelley
机构
[1] University of Nevada,Computer Science and Engineering Department
[2] Reno (UNR),undefined
来源
Neural Processing Letters | 2016年 / 43卷
关键词
Spiking neuron network; Synaptic plasticity; Polychronization; Unsupervised learning; Classification; STDP; Early classification;
D O I
暂无
中图分类号
学科分类号
摘要
This paper addresses the problem of encoding and classifying spatio-temporal patterns, which are typical for human actions or gestures. The proposed method has the following main contributions: (i) it requires a very small number of training examples, (ii) it accepts variable sized input patterns, (iii) it is invariant to scale and translation, and (iv) it enables early recognition, from only partial information of the pattern. The underlying representation employed is a spiking neural network with axonal conductance delay. We designed a novel approach for mapping spatio-temporal patterns to spike trains, which are used to stimulate the network. The pattern features emerge in the network as a result of this stimulation in the form of polychronous neuronal groups, which are used for classification. The proposed method is validated on a set of gestures representing the digits from 0\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$0$$\end{document} to 9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$9$$\end{document}, extracted from video data of a human drawing the corresponding digits. The paper presents a comparison with several other standard pattern recognition approaches. The results show that the proposed approach significantly outperforms these methods, it is invariant to scale and translation, and it has the ability to recognize patterns from only partial information.
引用
收藏
页码:327 / 343
页数:16
相关论文
共 67 条
  • [1] Arulampalam MS(2002)A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking IEEE Trans Signal Process 50 174-188
  • [2] Maskell S(2004)The writer independent online handwriting recognition system frog on hand and cluster generative statistical dynamic time warping IEEE Trans Pattern Anal Mach Intel 26 299-310
  • [3] Gordon N(2013)Categorization and decision-making in a neurobiologically plausible spiking network using a stdp-like learning rule Neural Netw 48 109-124
  • [4] Clapp T(1997)A state-based approach to the representation and recognition of gesture IEEE Trans Pattern Anal Mach Intel 19 1325-1337
  • [5] Bahlmann C(2011)Libsvm: a library for support vector machines ACM Trans Intel Syst Technol 2 27-3220
  • [6] Burkhardt H(2010)Deep, big, simple neural nets for handwritten digit recognition Neural Comput 22 3207-2072
  • [7] Beyeler M(2009)A new algorithm and system for the characterization of handwriting strokes with delta-lognormal parameters IEEE Trans Pattern Anal Mach Intel 31 2060-868
  • [8] Dutt ND(2009)A novel connectionist system for unconstrained handwriting recognition IEEE Trans Pattern Anal Mach Intel 31 855-1554
  • [9] Krichmar JL(2004)A comprehensive foundation Neural Netw 2 2004-147
  • [10] Bobick AF(2006)A fast learning algorithm for deep belief nets Neural Comput 18 1527-28