期刊名称：International Journal of Advanced Computer Science and Applications(IJACSA)
印刷版ISSN：2158-107X
电子版ISSN：2156-5570
出版年度：2019
卷号：10
期号：11
页码：68-74
出版社：Science and Information Society (SAI)
摘要：Recognition of hand gesture is one of Human PCs
most growing interfaces. In most vision-based signal recognition
system, the initial phase is hand detection and separation.
Because the hands are linked to a variety of day by day, local
work experiences both extraordinary changes in the illumination
and the innate unbroken appearance of the hand. In order to
address these issues, we suggest another 2D hand position
software that can be seen as a combination of multi-feature hand
proposal generation and cascading neural system network
characterization (CCNN). When considering various luminances
we select color, Gabor, Hoard and Filter to separate the skin and
produce a hand proposal. Therefore, we are selling a cascaded
CNN that holds the deep setting information between the
proposals. A mix of some datasets, including a few Oxford Hands
Datasets, VIVA Hand Recognition, and Egohands Datasets, is
tested as the positive example and image patch Net 2012, FDDEB
dataset as a bad example; the proposed Multi-Feature Directed
Cascaded CNN (MFS-CCNN) strategy. Aggressive results are
achieved by the technique proposed. Our average sample dataset
accuracy is considerably inferior to DPM. With an average of
43.55 and 51.78 percent accuracy, our CCNN and MFS-CCNN
model perform DPM. Average accuracy of the CCNN model in a
combined test set is 9.16% higher than the SSD model. Still, our
model is faster than a DPM based on the statistical performance.
