this is for holding javascript data
Tash Diaz added Aside_from_using_EMG_signals__.tex
about 8 years ago
Commit id: b7578e61542f0da2ade4f75c9a7ab458b433ffd9
deletions | additions
diff --git a/Aside_from_using_EMG_signals__.tex b/Aside_from_using_EMG_signals__.tex
new file mode 100644
index 0000000..4453a8e
--- /dev/null
+++ b/Aside_from_using_EMG_signals__.tex
...
Aside from using EMG signals as computer inputs for game control, a notable study performed by Schuuurink et al. \cite{Schuurink_2008} applied the said technology for measuring the user's engagement in the game. It has been learned that effects of sound and dynamics in serious gaming have shown a significant influence on the affective appraisal of the environment.
\subsection{Related Studies on Machine Learning in processing EMG based gestures}
Several studies related to EMG pattern recognition have been conducted over the past decades. These techniques have been used to analyze EMG signals which have been complex to recognize due to large variations in signals. In a study conducted by Liu et al. \cite{Liu_2007}, a novel EMG classifier called cascaded kernel learning machine (CKLM) was proven to be effective, achieving a high recognition rate of 93.54\%. The study employed a cascaded architecture of kernel learning machines including the General Discriminant Analysis (GDA), and the support vector machine (SVM) which offers classification performance that matches or exceeds other classifiers and does so in a computationally efficient manner \cite{Oskoei_2008}.
In a study conducted by Yoshikawa et al. \cite{Yoshikawa_2006}, a real-time hand motion estimation was conducted by using EMG signals with SVMs. The study involved an experiment wherein a subject was required to perform seven hand motions sequentially for 10 sessions. High accuracy is maintained through nine of the sessions, one exhibiting a performance of 94.17\% accuracy.
In a similar study of real-time muscle estimation was conducted by Lozito et al. \cite{Lozito_2015}. In this work, a real-time embedded implementation for a biomechanical model of the leg was created. The neural estimator implemented in the model allowed two configurations of the model: Precision oriented and Performance oriented. According to Hiraiwa et al. \cite{Hiraiwa}, neural networks make EMG pattern recognition much easier and more efficient. Their study was able to successfully accomplish the classification of the 5 finger movements based on the recognition of their patterns.
A notable method in decoding individual flexion and extension movements of each finger has been presented by Tenore et al. \cite{Tenore_2009}. The study was able to decode the finger movements performed by a transradial amputee with greater then 90\% accuracy. This can be achieved by using a neural network classifier trained on a waveform length feature. Further decoding accuracy from a transradial amputee and able-bodied subjects show no statistically significant difference.
On the topic of EMG signal processing for simultaneous myoelectric control, the study by Nielsen et al. \cite{Nielsen_2009} would be noteworthy. Extracted neural-control information from a multi-channel surface electomyographic signal are read as input and then used to train a multilayer perceptron (MLP) neural network.
