Download - Academy Publisher

others through the selective emotional music, with a total
of Record a 25-day EMG physiological signals whose
signal sampling frequency is 32Hz.
This paper uses a quadratural, compact Daubechies
wavelet as the base function for six-scale decomposition
of the EMG physiological signal data each day. And
extract the maximum and minimum values composition
vector of each layer in wavelet decomposition as the
feature vector of the surface EMG signal vector,
constituting a 14-dimensional feature vector. The EMG
signals Waveform of joy and wavelet transform
coefficients Wf (a, b) in different scales are shows as fig.3.
For emotional joy pattern recognition next
procedure, Table I shows the maximum value of 6 level
wavelet decomposition in five experiments.
TABLE I. the maximum of typical surface EMG joy signal
joy
a6 d6 d5 d4 d3 d2 d1
1 137.9 29 11.99 14.38 16.67 13.76 7.097
2 130.6 24.51 10.11 40.3 27 50.58 27.75
3 209.8 20.83 40.21 18.28 18.53 17.68 18.77
4 133.4 24.95 11.21 23.47 16.94 18.01 9.231
5 91.27 21.91 9.811 13.92 8.421 8.224 8.942
Then a three-layer BP neural network is used in this
paper. Input nodes number is 14 and the output nodes
number is 4, which represents four kinds of emotional
states, respectively, joy(1000), anger(0100), sadness(0010)
and pleasure(0001). For how number of the nodes
selection in the middle hidden layer, experiments show
that the hidden layer nodes have a significant impact for
the performance of neural networks. If we select too few
nodes, each category can not be separated by the network,
and if too many nodes, the operation is too big, there
maybe "over learning". Therefore system performance
and efficiency must be taken into comprehensive
consideration to determine the hidden layer nodes. In this
study, after many experiments comparison, finally we
select 14 as the hidden layer nodes and the effect is quite
good. The training sample set is closely related to network
performance. To design a good set of training samples, it
is necessary to select the sample size and the quality of the
sample is important. That is, the determination of the
samples number and sample selection organization is
important. In this paper, data of 19 days are selected as the
training set. Through experimental selection and
comparison, the remaining data of six days are selected as
test set. Learning rate has a great impact on network
performance too. The experiments show that when the
learning rate is 0.01 and the precision control parameter
will under 0.01.
Then we get the result of emotion recognition by the
use of BP neural network. After many experiment, the
recognition rate can reach more than 83.33%. We can see
that the emotion recognition using the BP neural network
is feasible. BP neural network for solving linear equations
used to achieve the classification makes the training time
greatly reduced.
IV. CONCLUTION
Emotion recognition has a promising development
prospect. Identifying the person's emotional state through
the physiological signal has drawn increasing attention.
This experiment introduce multi-scale decomposition
wavelet of EMG signals by wavelet transform and extract
the maximum and minimum of wavelet decomposition
coefficients to construct signal feature vector to present
the original EMG. Then enter it into the standard BP
neural network classifier for emotion recognition. The
kind of classify is able to detect and identify the surface
EMG of four kinds of emotions, joy, anger, sadness and
pleasure. Compared to the classical classifier, the emotion
recognition has a better classification effect, higher
recognition rate and better robustness. Experimental
results show that the surface EMG feature extraction
based on wavelet transform and emotional type
recognition method using BP neural network as a
classification tool is feasible and effective in application
of emotion recognition.
REFERENCES
[1] Luo Senlin, Pan Limin. Affective Computing Theory and
Technology [J]. Systems Engineering and Electronics,
2003,25 (7) :905-909.
[2] Picard R. Affective Computing [M]. England: MIT Press,
1997:68.
[3] Kim KH, Bang SW, Kim S R. Emotion recognition system
using short-term monitoring of physiological signals [J].
Medical Boilogical Engineering Computing, 2004, 42
(3) ,419-427.
[4] Wagner J, Kim J, AndréE. From physiological signals to
emotions: implementing and comparing selected methods
for feature extraction and classification [J]. IEEE
International Conference on Multimedia & Expo.New
York: IEEE Computer Society Press, 2005:940 -943.
[5] Yang Rui request, Liu Guangyuan. BPSO four
physiological signals based on the emotional state
identification [J]. Computer Science, 2008,35(03),137-
139.
[6] Niu Xiao-wei, LIU Guang-yuan. Physiology signals based
on genetic algorithm emotion recognition [J]. Southwest
University (Natural Science Edition), 2007, (9) ,134-138.
[7] Wu Fang-liang, Shi Zhong Kun, Yang Xiang-hui, et al.
Based on BP neural network with LM algorithm submarine
sonar self-noise prediction [J]. Chinese shipbuilding,
2006,47 (3): 45 - 50.
[8] Kun, Wang Zhizhong. Using LM algorithm is an improved
BP network in Surface EMG Signal in the applied research
[J]. R & D on the, 2005,29 (6) :399-401.
229