This paper describes CMVF, a new framework for indexing multimedia data
using multiple data properties combined with a neural network. The goal of
this system is to allow straightforward incorporation of multiple image feature
vectors, based on properties such as colour, texture and shape, into a single
low-dimensioned vector that is more effective for retrieval than the larger
individual feature vectors. CMVF is not constrained to visual properties, and
can also incorporate human classification into the features to further enhance
retrieval effectiveness. The analysis in this paper concentrates on CMVF's perfo
rmance
on images, examining how the incorporation of extra features into the indexing a
ffects
both performance and effectiveness, and demonstrating that CMVF's effectiveness 
is
robust against various kinds of common image distortions and initial(random)
configuration of neural network.