Muscle Computer Interfacing (Myoelectric Control)
The loss of the human forearm is a major disability that profoundly limits the everyday capabilities and interactions of individuals with upper-limb amputation (Kuiken et al., 2009). The interaction capability with the real-world can be restored using myoelectric control (Englehart & Hudgins, 2003; Hudgins, Parker, & Scott, 1993), where the electromyogram (EMG) signals generated by the human muscles are used to derive control commands for powered upper-limb prostheses. In the following videos, we are demonstrating the concept of myoelectric control in real-time. These are simply demonstrations associated with few of the papers we published in this field.
The loss of the human forearm is a major disability that profoundly limits the everyday capabilities and interactions of individuals with upper-limb amputation (Kuiken et al., 2009). The interaction capability with the real-world can be restored using myoelectric control (Englehart & Hudgins, 2003; Hudgins, Parker, & Scott, 1993), where the electromyogram (EMG) signals generated by the human muscles are used to derive control commands for powered upper-limb prostheses. In the following videos, we are demonstrating the concept of myoelectric control in real-time. These are simply demonstrations associated with few of the papers we published in this field.
Related Paper:
R. N. Khushaba, S. Kodagoda, Dikai Liu, and G. Dissanayake, "Electromyogram (EMG) based Fingers Movement Recognition Using Neighborhood Preserving Analysis with QR-Decomposition", Seventh International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP), pp. 1-6, 2011.
This is an initial testing session for a myoelectric control system (finger movement classification). It includes the classification of 10 classes of individual and combined fingers movements with EMG data collected by Delsys Bagnoli system (8 channels). TDAR features were extracted and projected with Fuzzy Neighborhood Preserving Analysis with QR-decomposition, while being classified with SVM. The user picks the testing classes in the first video while being instructed to perform random movements in the second video. More details about implementation will be given after publication of the associated paper in ISSNIP 2011.
R. N. Khushaba, S. Kodagoda, Dikai Liu, and G. Dissanayake, "Electromyogram (EMG) based Fingers Movement Recognition Using Neighborhood Preserving Analysis with QR-Decomposition", Seventh International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP), pp. 1-6, 2011.
This is an initial testing session for a myoelectric control system (finger movement classification). It includes the classification of 10 classes of individual and combined fingers movements with EMG data collected by Delsys Bagnoli system (8 channels). TDAR features were extracted and projected with Fuzzy Neighborhood Preserving Analysis with QR-decomposition, while being classified with SVM. The user picks the testing classes in the first video while being instructed to perform random movements in the second video. More details about implementation will be given after publication of the associated paper in ISSNIP 2011.
Muscle Computer Interfacing - Driver Distraction Avoidance
Related Paper
R. N. Khushaba, S. Kodagoda, D. Liu, and G. Dissanayake, "Muscle Computer Interfaces for Driver Distraction Reduction", Computer Methods and Programs in Biomedicine, vol. 110, no. 2, pp. 137-149, 2013.
A new control scheme is presented in this research based on finger force actions that would allow the control of external devices without the need to remove the hand/s from the item being held, or without the need to look at a screen or button console. It can be used for controlling in-car entertainment, phone or other devices without the need to remove hands from the steering wheel, or similarly for motorbike riders or cyclists to be able to control devices without taking their hands from the handlebars. In the proposed control scheme, there is no need to modify the steering wheel or handlebars, as all the hardware can be attached to the user in the form of an arm-band (though am showing an actual set of sensors attached here to prove the concept).
Related Paper
R. N. Khushaba, S. Kodagoda, D. Liu, and G. Dissanayake, "Muscle Computer Interfaces for Driver Distraction Reduction", Computer Methods and Programs in Biomedicine, vol. 110, no. 2, pp. 137-149, 2013.
A new control scheme is presented in this research based on finger force actions that would allow the control of external devices without the need to remove the hand/s from the item being held, or without the need to look at a screen or button console. It can be used for controlling in-car entertainment, phone or other devices without the need to remove hands from the steering wheel, or similarly for motorbike riders or cyclists to be able to control devices without taking their hands from the handlebars. In the proposed control scheme, there is no need to modify the steering wheel or handlebars, as all the hardware can be attached to the user in the form of an arm-band (though am showing an actual set of sensors attached here to prove the concept).
Related Paper:
R. N. Khushaba, S. Kodagoda, Dikai Liu, and G. Dissanayake, "Electromyogram (EMG) based Fingers Movement Recognition Using Neighborhood Preserving Analysis with QR-Decomposition", Seventh International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP), pp. 1-6, 2011.
Real-Time Myoelectric Control: Finger movement recognition using 2 electrodes, 10 classes. A new feature extraction method is being tested that characterizes the EMG patterns in terms of moments representing amplitude, time-scale, complexity, sparsity measure and spatial information allowing very fast neural information discovery. However, please note two important issues: transients were not included in training so there are some slight errors because of that. Additionally, the effect of speed of movement is also clear here. If you take longer time in implementing the movement then the classifier is not able to infer the class during transients that are taking some time. Suggestions are to include transients in training and make it fast transitions (any other suggestions ?). Window size 125 msec upon which decisions are made and smoothed with MV and classifier type is support vector machine (SVM).
R. N. Khushaba, S. Kodagoda, Dikai Liu, and G. Dissanayake, "Electromyogram (EMG) based Fingers Movement Recognition Using Neighborhood Preserving Analysis with QR-Decomposition", Seventh International Conference on Intelligent Sensors, Sensor Networks and Information Processing (ISSNIP), pp. 1-6, 2011.
Real-Time Myoelectric Control: Finger movement recognition using 2 electrodes, 10 classes. A new feature extraction method is being tested that characterizes the EMG patterns in terms of moments representing amplitude, time-scale, complexity, sparsity measure and spatial information allowing very fast neural information discovery. However, please note two important issues: transients were not included in training so there are some slight errors because of that. Additionally, the effect of speed of movement is also clear here. If you take longer time in implementing the movement then the classifier is not able to infer the class during transients that are taking some time. Suggestions are to include transients in training and make it fast transitions (any other suggestions ?). Window size 125 msec upon which decisions are made and smoothed with MV and classifier type is support vector machine (SVM).
Recognizing Fingers Movements Across Multiple Limb-Positions:
This is a real-time demo on using the time-domain spectral moments described in the following paper below to implement finger movement recognition. As can be seen, the proposed surface EMG pattern recognition system is successful at recognizing almost all of these movement. It should be mentioned here that the further training for the subjects on the proposed EMG system can enhance the results. This demo was made based on EMG sensors only, i.e., no accelerometers were utilized at all. The paper targeted 8 classes of simple hand movements, but the demo below was made to test if we can classify fingers movements as well using this approach.
Related Paper
[1] R. N. Khushaba, L. Shi, and S. Kodagoda, "Time-Dependent Spectral Features for Limb Position Invariant Myoelectric Pattern Recognition", in Proc. Int. Symp. on Communications & Information Technologies (ISCIT2012), Gold Coast, pp. 1020-1025, 2012.
[2] R. N. Khushaba, Maen Takruri, Jaime Valls Miro, and Sarath Kodagoda, "Toward Limb Position Invariant Myoelectric Pattern Recognition Using Local-Global Time-Dependent Spectral Features", Submitted for publication, ... an extension of the above.
