Download - Academy Publisher
Download - Academy Publisher
Download - Academy Publisher
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
ISBN 978-952-5726-09-1 (Print)<br />
Proceedings of the Second International Symposium on Networking and Network Security (ISNNS ’10)<br />
Jinggangshan, P. R. China, 2-4, April. 2010, pp. 227-229<br />
Emotion Recognition of EMG Based on BP<br />
Neural Network<br />
Xizhi Zhu<br />
School of Physical & Electronics Engineering, Taizhou University, Taizhou, China<br />
zxztzc@126.com<br />
Abstract— This paper presents the emotion recognition of<br />
BP neural network classifier. Experiment analyzes wavelet<br />
transform of surface Electromyography (EMG) to extract<br />
the maximum and minimum multi-scale wavelet coefficients<br />
firstly. And then we enter the two kinds of structural<br />
feature vector classifier for emotion recognition. The<br />
experimental results show that both classifiers shows a<br />
effective recognition on the four kinds of emotions, joy,<br />
anger, sadness, pleasure, etc. Also the experimental results<br />
show that feature vector extracted by wavelet transform can<br />
characterize emotional patterns, and it has a stronger<br />
emotional recognition effect than the traditional classify<br />
method.<br />
Index Terms—Surface Electromyography(EMG)Signal;<br />
Emotional Recognition; Wavelet Transform; BP Neural<br />
Network<br />
I. INTRODUCTION<br />
The study of emotion recognition has an important<br />
significance in understanding human emotions in the role<br />
of human intelligence. In daily life, human intelligence<br />
not only shows in the normal rational thinking and logical<br />
reasoning ability, but also in the normal emotional<br />
capabilities. In computer science, this ability to promote<br />
the establishment of a friendly man-machine interface is<br />
of great significance. As the deepening of affective<br />
computing, the request for emotion recognition<br />
[1, 2]<br />
technology will be correspondingly enhanced.<br />
The MIT media experiment team led by Professor<br />
Picard collected a 20-day physiological signal from eight<br />
kinds of emotions of an actor when performing<br />
deliberately, to extract the value of demographic<br />
characteristics using way of Fisher Projection [2] . At last<br />
they get 83% of the emotion recognition rate [3] . In<br />
Augsburg University, Germany, Johannes Wagner and<br />
others got approximately 80% of the emotion recognition<br />
rate on a subject's evoked in the music under joy, anger,<br />
sadness, pleasure of the four kinds of EMG signal by the<br />
use of physical K-nearest neighbor, linear discriminant<br />
function and multi-layer in 25 days [4] . Among them,<br />
Yang Ruiqing and Liu Guangyuan led the feature<br />
selection by the use of discrete binary particle swarm<br />
optimization (BPSO), proving that the method of BPSO<br />
on physiological signal for feature selection is feasible [5] .<br />
Also Niu Xiaowei and Liu Guangyuan selected the most<br />
representative characteristics of the corresponding optimal<br />
combination of emotional states by using the genetic<br />
algorithms, proving that the optimization problem, the<br />
genetic feature selection algorithm being used to identify<br />
© 2010 ACADEMY PUBLISHER<br />
AP-PROC-CS-10CN006<br />
227<br />
the choice of the combinational optimal portfolio of the<br />
physiological signal recognition optimal emotional<br />
characteristics, is feasible [6] .<br />
In this paper, the method of Wavelet Transform is<br />
used in surface EMG aiming for non-stationary features of<br />
surface EMG signal in order to extract more effective,<br />
reliable, robust signal characteristics. This will help<br />
improve the recognition rate of the surface EMG. Also<br />
this paper uses the surface EMG signal with objective data<br />
for six-scale decomposition of surface EMG with the<br />
method of wavelet transform and extract the maximum of<br />
multi-scale wavelet coefficients, constructing 14-<br />
dimensional feature vector, then we input into the BP<br />
neural network for emotion pattern recognition.<br />
II. RECOGNITION METHODS OF EMOTIONAL MOTION<br />
BP neural network is a multi-level error feedback<br />
network proposed by Rumelhart and Mc Clelladn in 1985.<br />
It uses the difference between the actual output and the<br />
desired output to correct the connection weight of the<br />
network and the threshold of each node in each layer<br />
from back to front. BP neural network contains an input<br />
layer, a middle layer (hidden layer) and an output layer.<br />
There is a full connectivity between the upper and lower<br />
layers and no connections between neurons in each layer.<br />
For the input signal, it needs to spread towards to hidden<br />
layer nodes and transformed by the function, then<br />
transmit the input signal of hidden layer nodes to the<br />
output layer nodes. Usually, the transfer function of BP<br />
neural network is Sigmoid Type differentiable function,<br />
which can achieve arbitrary non-linear mapping between<br />
the input and output, so BP network has been widely<br />
applied in pattern recognition, function approximation<br />
and other areas [7, 8] .<br />
The three nodes of the BP network is represented as:<br />
Input layer Hidden layer Output layer<br />
Figure 1. BP neural network structure