Article Preview
TopIntroduction
An accurate, reliable and clinically reasonable method to detect critical gait events is required towards improving the human locomotion by restoring the functionality of weak or paralyzed lower extremity muscles. Thus, gait event detection received considerable attention based on a combination of hardware and sophisticated software algorithms. Various signal sources are used for gait event detection including foot switches (Ng, 1997), accelerometers (Crago et al., 1996), angle velocity’s (Lyons et al., 2002) and nerve’s sensors (Hansen, 2001) among others.
When foot switches are used, an appropriate positioning is required to obtain a reliable detection for each subject, especially those suffered from pathological gait (Smith et al., 2002). A frequent maintenance is required since these sensors suffer from the initial contact impact. Information content is limited, unless sophisticated algorithms like fuzzy inference systems (Skelly et al., 2001) are employed, or there is an increase in the number of the attached foot switches (Smith et al., 2002). Accelerometers and combinations of accelerometers with angle velocity’s sensors are used for the detection of the Initial Contact (IC) event in cases of the lower limb pathology, called Drop-Foot Syndrome (Lyons et al., 2002). Various devices are developed such as accelerometers and gyroscopes based on recent advances in miniaturization, which were used for offline (Jasiewicz et al., 2006) and online gait event detection (Hanlon & Anderson, 2006). Gonzalez et al (2010) introduced a heuristic rule-based algorithm for the real-time detection and timing of the Initial Contact (IC) event between two gait cycles, processing vertical and antero-posterior accelerations registered at the lower trunk.
The aforementioned sensor choices presuppose the placing of sensors onto or into (implantable sensors) the patient limbs with a number of issues still pending, such as sensor maintenance, donning, cosmetic, etc. Moreover, gait event detection is conducted in the laboratory, where the force that a patient applies during walking is measured using a force plate. Gait event detection is performed by processing kinematic data obtained with video-cameras and markers, placed on the patient during human locomotion. Although force plates and kinematic data processing methods are often considered to be the “gold standard” for determining gait events, these are laboratory methods and cannot be used for rehabilitation purposes.
Another source for gait event detection and rehabilitation control is an electromyographic (EMG) signal. The use of EMG for switching either the initiation or ending an action for the control of upper limb Functional Electrical Stimulation (FES) systems was reported in (Nathan, n. d.; Hart, 1998). Further elaborated control schemes, such as the use of a time-delayed adaptive neural network (TDANN) to predict movements of the shoulder and elbow for an upper limb FES system (Au & Kirsch, 2000) or a fuzzy inference system (FIS) extracting information from the EMG signal to operate an externally powered prosthesis (Park & Lee, 1998), were proposed.
Gait event detection using EMG signals from voluntarily controlled muscles was proposed by two research groups. Kordylewski and Graupe (2001) acquired surface EMG signals from upper trunk muscles to predict intended lower extremity movements for reciprocal walking or postural corrections for static standing using neural networks. Lauer et al (2005) tested the feasibility of gait event detection by reading the intramuscular EMG from a lower limb muscle, called quadriceps, using a Fuzzy and a Neuro-Fuzzy network, in children with cerebral palsy. The results were very promising towards using data from lower limb muscles EMGs for the development of gait event detection techniques and rehabilitation control.
In addition, the EMG signal is attractive for gait phase detection and rehabilitation control for the following reasons. Proximal lower extremity muscle activity occurs in a repeatable way with respect to the gait cycle (Hof et al., 2002). The technology achieved fully implantable tiny recording electrodes (Smith et al., 1998), instead of using other implantable larger and more complex sensor devices that require extensive maintenance and donning.