Information Technology Reference
In-Depth Information
linguistic report about the quality of the gait in terms of homogeneity and symme-
try. This type of reports could be used to analyze the evolution of the human gait
after a recovery treatment and also for preventing falls in elderly people.
The remainder of this work is organized as follows. Section 11.2 presents the
human gait modeling problem and our proposal for tackle it. Section 11.3 describes
how to automatically generate a linguistic report on the quality of the gait based on
the features obtained by our modeling system. Finally, Section 11.4 draws some
conclusions and introduces some future directions in this research line.
11.2
Gait Modeling
11.2.1
Proposal
Human gait modeling consists of studying the biomechanics of this human move-
ment aimed at quantifying factors governing the functionality of the lower extrem-
ities. Gait is a complex integrated task which requires precise coordination of the
neural and musculoskeletal system to ensure correct skeletal dynamics [15]. There-
fore, its analysis can help in the diagnosis and treatment of walking and movement
disorders, identification of balance factors, and assessment of clinical gait interven-
tions and rehabilitation programs [8, 11].
The gait cycle is a periodical phenomenon which is defined as the interval be-
tween two successive events (usually heel contact) of the same foot [5]. It is char-
acterized by a stance phase (60% of the total gait cycle), where at least one foot is
in contact with the ground, and a swing phase (40% of the total gait cycle), during
which one limb swings through the next heel contact (see Fig. 11.1). These phases
can be quite different between individuals but when normalized to a percentage of
the gait cycle they maintain close similarity, indicating the absence of disorders [12].
We base on the accelerations produced during the human gait cycle. We use
a expert knowledge based fuzzy finite state machine (FFSM) as a modeling tool,
which has also been used to extract relevant features for the authentication purpose
[14]. The main advantage of using this tool is its flexibility when dealing with the
variations in both amplitude and states time span. The fuzziness of the model allows
us to handle imprecise and uncertain data which is inherent to real world phenomena
in the form of fuzzy if-then rules. Moreover, the use of linguistic terms makes easier
its interpretation and does not require high computational cost thanks to the lack of
a learning process. Nevertheless, there exists the possibility of making use of an
automatic machine learning technique to design the main elements of the FFSM as
explained in [1].
We attached a smartphone equipped with a three-axial accelerometer to a belt,
centered in the back of the subject. The smartphone executes an application which
provide us with the dorso-ventral acceleration ( a x ), the medio-lateral acceleration
( a y ), and the antero-posterior acceleration ( a z ) at each instant of time. In this con-
tribution, we only use a x and a y because a z has to do with the walking speed and
this speed can vary for the same person. Therefore, every record contained the three
 
Search WWH ::




Custom Search