General Perspectives on Electromyography Signal Features and Classifiers Used for Control of Human Arm Prosthetics

General Perspectives on Electromyography Signal Features and Classifiers Used for Control of Human Arm Prosthetics

DOI: 10.4018/978-1-5225-7359-3.ch001
(Individual Chapters)
No Current Special Offers


Physically handicapped people encounter various kinds of obstacles and difficulties in their daily lives due to the restricted ability of motion. Assistive technologies represent a crucial challenge of scientific studies to overcome such an issue of reducing quality of life. Assistive devices such as wheelchairs, orthoses, and prostheses are designed and built to contribute rehabilitation progress and to regain lost functions. Although human body parts have intricate forms and functions, artificial devices and components integrating to the body are anticipated to compensate the fundamental functions related to user's demands. Upper- or lower-arm amputations also result in severe cosmetic matters. However, what is more important and obtrusive is the loss of primary functions including manipulating and grasping the objects besides the locomotor tasks which are performed by the human body during daily activity.
Chapter Preview

Fundamental Aspects Of Emg

EMG is the electrical activity of skeletal muscles (Basmajian & Deluca, 1985). It represents the summation of the muscle action potentials which cause the contraction of muscle fibers. Recorded EMG data by means of electrodes are amplified and filtered to eliminate the motion artifacts, as well as the environment and device related noises. Rejection of ambient influences on natural muscle activation improves the accuracy and usability of EMG signals. One of the most widely usage of EMG signals is to control the myoelectric-based prosthetics which are used by amputated people. Control scheme for EMG-driven human arm prosthetics includes a sequential series of signal processing (Figure 1).

Figure 1.

Control scheme of multifunctional human arm prosthetics


A condensed and clear control signal is needed to control the EMG-based prosthetics. In order to reduce calculation and to provide stability of signal, EMG data are scanned by sliding segmented windows (Figure 2). Because the raw (amplified+ pre-processed) EMG signal contains a huge burden of data, this signal is needed to be represented in a concise, but accurate ways. Widely used time domain features extracted from signals includes mean absolute value (MAV), root mean square (RMS), Willison amplitude (WAMP), waveform length (WL), variance of EMG (VAR), simple square integral (SSI), zero crossing (ZC) and integrated EMG (IEMG) (Phinyomark et al., 2013). In frequency domain, mean frequency, median frequency, peak frequency, mean power, total power, and spectral power features are commonly preferred (Phinyomark et al., 2013).

Figure 2.

The basic representation of sliding windows


Key Terms in this Chapter

Pattern Recognition: A machine learning process which identifies the pattern of physical systems using data belong to investigated systems.

Rehabilitation: A series of therapy to make injured or amputated people regained lost skills or functions.

Human Arm Prostheses: Assistive devices which enable to perform lost functions of human arm due to upper or lower arm amputations.

Surface Electromyography: A type of electromyography signal recording method carrying out by means of adhering electrodes to skin surface.

Assistive Technology: A branch of technology is used to regain the lost functions of human body parts.

Feature Extraction: A method to obtain meaningful and clear data of a signal.

Feature Classification: A pattern recognition technique that is used to categorize a huge number of data into different classes.

Complete Chapter List

Search this Book: