Relationship between grasping force and features of single-channel intramuscular EMG signals
Introduction
Electromyographic (EMG) signals are widely used for the control of prosthetic devices (myoelectric control, e.g. Otto Bock DMC Plus®). The EMG signal features are typically used for predicting either the intended limb movement (Hargrove et al., 2007, Farrell and Weir, 2008) or the amount of force required to execute a task (Duque et al., 1995, Hoozemans and Van Dieën, 2005). The main advantage of myoelectric prosthesis over other systems, e.g. body-powered prostheses, is that myoelectric control is close to the physiological experience of limb control. Non-invasive EMG recordings (surface EMG, sEMG) are most commonly used for this purpose. For example, grasping force can be predicted from sEMG because of the monotonic relationship between sEMG amplitude and force, which can be linear (Inman et al., 1952, Hoozemans and Van Dieën, 2005) or non-linear (Zuniga et al., 1970, Herzog et al., 1998, Liu et al., 1999).
The use of intramuscular EMG (iEMG) for myoelectric control has been less explored due to technical difficulties. However, reliable, implantable electrodes have been proposed recently (Weir et al., 2005, Farina et al., 2008b). Thus iEMG interfaces for myoelectric control may be chronically implanted and may provide more stable and more selective recordings than sEMG. Furthermore, the use of iEMG will not require appropriate control signal sites to be on superficial muscles. Though, the greater selectivity of iEMG with respect to sEMG may be a disadvantage for the control since the signal may provide local, rather than global information on the intensity of muscle activity.
Grasping force is one of the main functions to investigate for applications in myoelectric prosthesis because of the highly important functional value of this task. When the number of DOFs is limited, a control command is typically predicted as the onset of muscle activity, and the amount of grip force and speed of the prosthetic device is estimated from the intensity of the EMG signal (e.g. Otto Bock DMC Plus® prosthesis). Nevertheless the capacity of iEMG to predict the power grasping force, in particular its linear relationship with force for proportional control, has not been investigated. In this study we investigated whether a highly selective recording (at the single motor unit level) was sufficient to estimate the grip force accurately.
The relationship between EMG features and force has been investigated since several decades (Inman et al., 1952, Bigland and Lippold, 1954, Perry and Bekey, 1981, Hof, 1997). The information on the intensity of muscle activity is usually extracted based on the smoothed integrated EMG (SIEMG) (Inman et al., 1952, Bouisset and Maton, 1972, Onishi et al., 2000) or a count of action potentials (Close et al., 1960, Bouisset and Maton, 1972). Due the high selectivity of the recording interface in this study, iEMG signals were processed in order to rely only on the modulation of the discharge rate, rather than the signal amplitude.
The multi-unit characteristics of iEMG has been investigated during single muscle isometric contraction mainly for understanding basic physiological processes, where, e.g. the recruitment strategies have been shown to relate to force (Freund et al., 1975, DeLuca et al., 1982). At a given time instant, the level of force is related to the total number of active motor units, from which a global discharge rate (GDR) can be estimated (total number of motor unit action potentials per unit time). With an intramuscular detection system, only few motor units are sampled and it is not known if these units are representative for the applied power grip force. Therefore the aim of this study was to quantify the linear correlation between grasping force and features of the iEMG and sEMG. We did not predict the force based on EMG signal; rather, we investigated whether power grip force was better described by a global measure of intensity (sEMG) or by a GDR in a local muscle area (iEMG), using linear correlation coefficients as measures.
Section snippets
Subjects
The experiments were conducted on 10 able-bodied human subjects (7 w/3 m; age range, 21–37 years, mean 26.9 years). The procedures were in accordance with the Declaration of Helsinki and approved by the Danish local ethical committee (approval no.: N-20080045). Subjects gave written informed consent prior to the experimental procedures. The subjects had no history of upper extremity or other musculoskeletal disorders.
Tasks and procedures
Subjects exerted handgrip forces with their right hand while seated
Results
Fig. 2 shows recorded signals of the grasping force for different profiles from one subject. For the entire group of profiles, the maximum grasping force (MGF) was found to be 233.4 ± 81.9 N. Thus 25 N (defined as the low-force range) and 50 N (defined as the high-force range) corresponded to 12.2 ± 5.2 and 24.4 ± 10.3% (mean ± SD) of the MGF, respectively.
Fig. 3A and B depict examples of sEMG and iEMG recordings from one subject. Fig. 3C and D show the sEMG and iEMG on a shorter time scale so that
Discussion
The results showed that features extracted from iEMG and sEMG correlate similarly with the grasping force. This result does not depend on the type of recording and processing used. Thus, a highly selective recording is representative of the applied power grasping force in the same manner as when using a more global measure of intensity.
Conclusion
This study assessed the correlation of grasping force with iEMG and compared it with sEMG. The results showed high correlation between features of a selective recording (iEMG) and force with no difference with respect to the correlation between sEMG and force. Thus a selective EMG recording is representative of the applied grasping force and can potentially be suitable for proportional control of prosthetic devices.
Acknowledgement
This study was supported by the Danish National Advanced Technology Foundation.
References (31)
- et al.
Effect of grip span on maximal grip force and fatigue of flexor digitorum superficialis
Appl Ergon
(1999) - et al.
Evaluation of handgrip force from EMG measurements
Appl Ergon
(1995) - et al.
EMG-force relation in dynamically contracting cat plantaris muscle
J Electromyogr Kinesiol
(1998) - et al.
Prediction of handgrip forces using surface EMG of forearm muscles
J Electromyogr Kinesiol
(2005) - et al.
Relation of human electromyogram to muscular tension
Electroencephalogr Clin Neurophysiol
(1952) - et al.
Dynamic muscle force predictions from EMG: an artificial neural network approach
J Electromyogr Kinesiol
(1999) - et al.
EMGLAB: an interactive EMG decomposition program
J Neurosci Methods
(2005) - et al.
Relationship between EMG signals and force in human vastus lateralis muscle using multiple bipolar wire electrodes
J Electromyogr Kinesiol
(2000) - et al.
Myoelectric signal processing for control of powered limb prostheses
J Electromyogr Kinesiol
(2006) - et al.
The relationship between force velocity and integrated electrical activity in human muscles
J Physiol
(1954)
Quantitative relationship between surface EMG and intramuscular electromyographic activity in voluntary movement
Am J Phys Med
Motor-unit action potential counts, their significance in isometric and isotonic contractions
J Bone Joint Surg Am
Single site electromyograph amplitude estimation
IEEE Trans Biomed Eng
Behaviour of human motor units in different muscles during linearly varying contractions
J Physiol
Amplitude cancellation of motor-unit action potentials in the surface electromyogram can be estimated with spike-triggered averaging
J Neurophysiol
Cited by (65)
A method for understanding and digitizing manipulation activities using programming by demonstration in robotic applications
2023, Robotics and Autonomous SystemsGrasping force prediction based on sEMG signals
2020, Alexandria Engineering JournalCitation Excerpt :Because of the complex structure of the hand, more degrees of freedom and various combinations of finger joint movements, the hand can achieve a variety of grasping movements. Literatures have studied the grasping maneuvers in daily life, and summarized six basic grasping modes: cylindrical, fingertip, hook, palmar, spherical and lateral [39,40]. Through the above six kinds of grasping modes, the human hand can complete most of the actions of grasping objects in daily life.
Hand gesture recognition based on motor unit spike trains decoded from high-density electromyography
2020, Biomedical Signal Processing and ControlCitation Excerpt :The motor unit discharges directly represents the neural drive to muscles from the spinal cord [42]. The contraction force of each motion can be estimated by the summed discharge rates [43], which are already extracted by the MUC approach, rather than by the global EMG amplitude, as in classic approaches. The summed discharge rates have been proved to outperform the RMS metrics in the estimation of muscle excitation [14].
The effect of arm position on classification of hand gestures with intramuscular EMG
2018, Biomedical Signal Processing and Control