Segmenting Mechanomyography Measures of Muscle Activity Phases Using Inertial Data.

Electromyography (EMG) is the standard technology for monitoring muscle activity in laboratory environments, either using surface electrodes or fine wire electrodes inserted into the muscle. Due to limitations such as cost, complexity, and technical factors, including skin impedance with surface EMG and the invasive nature of fine wire electrodes, EMG is impractical for use outside of a laboratory environment. Mechanomyography (MMG) is an alternative to EMG, which shows promise in pervasive applications. The present study used an exerting squat-based task to induce muscle fatigue. MMG and EMG amplitude and frequency were compared before, during, and after the squatting task. Combining MMG with inertial measurement unit (IMU) data enabled segmentation of muscle activity at specific points: entering, holding, and exiting the squat. Results show MMG measures of muscle activity were similar to EMG in timing, duration, and magnitude during the fatigue task. The size, cost, unobtrusive nature, and usability of the MMG/IMU technology used, paired with the similar results compared to EMG, suggest that such a system could be suitable in uncontrolled natural environments such as within the home.

Adapting myoelectric control in real-time using a virtual environment.

BACKGROUND: Pattern recognition technology allows for more intuitive control of myoelectric prostheses. However, the need to collect electromyographic data to initially train the pattern recognition system, and to re-train it during prosthesis use, adds complexity that can make using such a system difficult. Although experienced clinicians may be able to guide users to ensure successful data collection methods, they may not always be available when a user needs to (re)train their device. METHODS: Here we present an engaging and interactive virtual reality environment for optimal training of a myoelectric controller. Using this tool, we evaluated the importance of training a classifier actively (i.e., moving the residual limb during data collection) compared to passively (i.e., maintaining the limb in a single, neutral orientation), and whether computational adaptation through serious gaming can improve performance. RESULTS: We found that actively trained classifiers performed significantly better than passively trained classifiers for non-amputees (P < 0.05). Furthermore, collecting data passively with minimal instruction, paired with computational adaptation in a virtual environment, significantly improved real-time performance of myoelectric controllers. CONCLUSION: These results further support previous work which suggested active movements during data collection can improve pattern recognition systems. Furthermore, adaptation within a virtual guided serious game environment can improve real-time performance of myoelectric controllers.

Robust Pattern Recognition Myoelectric Training for Improved Online Control within a 3D Virtual Environment.

It has been shown that maintaining a neutral arm position during collection of pattern recognition training data for myoelectric prosthesis control results in high offline classification accuracies; however, that precision does not translate to real-time applications, when the arm is used in different positions. Previous studies have shown that collecting training data with the arm in a variety of positions can improve pattern recognition control systems. In this work, we extended these findings to real-time myoelectric control in an immersive testing environment using virtual reality. We show that collecting training data for a pattern recognition algorithm under dynamic conditions, where the user moves their arm, significantly improves control efficiency and achievement of testing metrics.

Comparing the Effects of Signal Noise on Pattern Recognition and Linear Regression-Based Myoelectric Controllers.

Myoelectric pattern recognition using linear discriminant analysis (LDA) classifiers has been a wellestablished control method for upper limb prostheses for many years. More recently, linear regression (LR) controllers have been proposed as an alternative solution due to their ability to control multiple degrees of freedom (DOF) simultaneously. The aim of this experiment was to compare the online performance of LDA and LR control systems under three electromyographic (EMG) signal conditions: baseline, noise in all channels, and noise in a single channel. To simulate the last two conditions, different levels of Gaussian noise were added to the EMG signals. Completion rate, path efficiency, dwelling time, and completion time were computed after virtual Fitts' Law tasks. While both controllers were significantly affected by the lowest noise levels, we found no significant differences between the controllers under the baseline and all-channel noise conditions. However, the LDA controller outperformed the LR controller in the single-channel noise condition. Therefore, while both controllers are comparable in most cases, the added complexity of simultaneous control affects an LR controller's performance under certain noise conditions. Based on these results, neither control system should be dismissed in future developments.

User intent prediction with a scaled conjugate gradient trained artificial neural network for lower limb amputees using a powered prosthesis.

Powered lower limb prostheses have the ability to provide greater mobility for amputee patients. Such prostheses often have pre-programmed modes which can allow activities such as climbing stairs and descending ramps, something which many amputees struggle with when using non-powered limbs. Previous literature has shown how pattern classification can allow seamless transitions between modes with a high accuracy and without any user interaction. Although accurate, training and testing each subject with their own dependent data is time consuming. By using subject independent datasets, whereby a unique subject is tested against a pooled dataset of other subjects, we believe subject training time can be reduced while still achieving an accurate classification. We present here an intent recognition system using an artificial neural network (ANN) with a scaled conjugate gradient learning algorithm to classify gait intention with user-dependent and independent datasets for six unilateral lower limb amputees. We compare these results against a linear discriminant analysis (LDA) classifier. The ANN was found to have significantly lower classification error (P<;0.05) than LDA with all user-dependent step-types, as well as transitional steps for user-independent datasets. Both types of classifiers are capable of making fast decisions; 1.29 and 2.83 ms for the LDA and ANN respectively. These results suggest that ANNs can provide suitable and accurate offline classification in prosthesis gait prediction.