Title: Fast and Accurate Gesture Spotting using Subgesture Reasoning and
Pruning of Unlikely Dynamic Programming Paths

Authors: Jonathan Alon, Vassilis Athitsos, and Stan Sclaroff

Date: June 3, 2005

Abstract:

Vision-based recognition of gestures in continuous video
streams can facilitate more natural human-computer interaction.
Gesture spotting is the challenging task of locating
the start and end frames of the video stream that correspond
to a gesture of interest, while at the same time rejecting
non-gesture motion patterns. This paper proposes a new
gesture spotting and recognition algorithm that is based on
the widely used continuous dynamic programming (CDP)
algorithm. Our first contribution is a pruning method that
allows the system to evaluate a relatively small number of
hypotheses compared to CDP. Pruning is implemented by
a set of model-dependent classifiers, that are learned from
training examples. In our experiments, the proposed CDP
with pruning was an order of magnitude faster compared
to the original CDP algorithm, and recognition accuracy
improved by 7%. The second contribution of the proposed
spotting algorithm is a subgesture reasoning process that
models the fact that some gesture models can falsely match
parts of other longer gestures. In our experiments, using the
proposed subgesture modeling improved recognition accuracy
by an additional 12%.