We propose refined composite multivariate multiscale fuzzy entropy (RCmvMFE).
The coarse-graining step of RCmvMFE uses variance (RCmvMFEclass="mathmlsrc">class="formulatext stixSupport mathImg" data-mathURL="/science?_ob=MathURL&_method=retrieve&_eid=1-s2.0-S0378437116305404&_mathId=si98.gif&_user=111111111&_pii=S0378437116305404&_rdoc=1&_issn=03784371&md5=6c2a18513051787f9e74709cdef27912" title="Click to view the MathML source">σ2class="mathContainer hidden">class="mathCode">) or mean (RCmvMFEclass="mathmlsrc">class="formulatext stixSupport mathImg" data-mathURL="/science?_ob=MathURL&_method=retrieve&_eid=1-s2.0-S0378437116305404&_mathId=si99.gif&_user=111111111&_pii=S0378437116305404&_rdoc=1&_issn=03784371&md5=a693ecbf6f72ed21af9613af4a0d15d0" title="Click to view the MathML source">μclass="mathContainer hidden">class="mathCode">).
The introduced fuzzy membership function significantly decreases the running time.
Our simulations demonstrate that RCmvMFEclass="mathmlsrc">class="formulatext stixSupport mathImg" data-mathURL="/science?_ob=MathURL&_method=retrieve&_eid=1-s2.0-S0378437116305404&_mathId=si98.gif&_user=111111111&_pii=S0378437116305404&_rdoc=1&_issn=03784371&md5=6c2a18513051787f9e74709cdef27912" title="Click to view the MathML source">σ2class="mathContainer hidden">class="mathCode"> and RCmvMFEclass="mathmlsrc">class="formulatext stixSupport mathImg" data-mathURL="/science?_ob=MathURL&_method=retrieve&_eid=1-s2.0-S0378437116305404&_mathId=si99.gif&_user=111111111&_pii=S0378437116305404&_rdoc=1&_issn=03784371&md5=a693ecbf6f72ed21af9613af4a0d15d0" title="Click to view the MathML source">μclass="mathContainer hidden">class="mathCode"> lead to more stable results.
RCmvMFEclass="mathmlsrc">class="formulatext stixSupport mathImg" data-mathURL="/science?_ob=MathURL&_method=retrieve&_eid=1-s2.0-S0378437116305404&_mathId=si98.gif&_user=111111111&_pii=S0378437116305404&_rdoc=1&_issn=03784371&md5=6c2a18513051787f9e74709cdef27912" title="Click to view the MathML source">σ2class="mathContainer hidden">class="mathCode"> and RCmvMFEclass="mathmlsrc">class="formulatext stixSupport mathImg" data-mathURL="/science?_ob=MathURL&_method=retrieve&_eid=1-s2.0-S0378437116305404&_mathId=si99.gif&_user=111111111&_pii=S0378437116305404&_rdoc=1&_issn=03784371&md5=a693ecbf6f72ed21af9613af4a0d15d0" title="Click to view the MathML source">μclass="mathContainer hidden">class="mathCode"> are less sensitive to the length of signals.