Adrian D. C. Chan

A Gaussian mixture model based classification scheme for myoelectric control of powered upper limb prostheses

This paper introduces and evaluates the use of Gaussian mixture models (GMMs) for multiple limb motion classification using continuous myoelectric signals. The focus of this work is to optimize the configuration of this classification scheme. To that end, a complete experimental evaluation of this system is conducted on a 12 subject database. The experiments examine the GMMs algorithmic issues including the model order selection and variance limiting, the segmentation of the data, and various feature sets including time-domain features and autoregressive features. The benefits of postprocessing the results using a majority vote rule are demonstrated. The performance of the GMM is compared to three commonly used classifiers: a linear discriminant analysis, a linear perceptron network, and a multilayer perceptron neural network. The GMM-based limb motion classification system demonstrates exceptional classification accuracy and results in a robust method of motion classification with low computational load.

Continuous myoelectric control for powered prostheses using hidden Markov models

This paper represents an ongoing investigation of dexterous and natural control of upper extremity prostheses using the myoelectric signal. The scheme described within uses a hidden Markov model (HMM) to process four channels of myoelectric signal, with the task of discriminating six classes of limb movement. The HMM-based approach is shown to be capable of higher classification accuracy than previous methods based upon multilayer perceptrons. The method does not require segmentation of the myoelectric signal data, allowing a continuous stream of class decisions to be delivered to a prosthetic device. Due to the fact that the classifier learns the muscle activation patterns for each desired class for each individual, a natural control actuation results. The continuous decision stream allows complex sequences of manipulation involving multiple joints to be performed without interruption. The computational complexity of the HMM in its operational mode is low, making it suitable for a real-time implementation. The low computational overhead associated with training the HMM also enables the possibility of adaptive classifier training while in use.

Wavelet Distance Measure for Person Identification Using Electrocardiograms

In this paper, the authors present an evaluation of a new biometric based on electrocardiogram (ECG) waveforms. ECG data were collected from 50 subjects during three data-recording sessions on different days using a simple user interface, where subjects held two electrodes on the pads of their thumbs using their thumb and index fingers. Data from session 1 were used to establish an enrolled database, and data from the remaining two sessions were used as test cases. Classification was performed using three different quantitative measures: percent residual difference, correlation coefficient, and a novel distance measure based on wavelet transform. The wavelet distance measure has a classification accuracy of 89%, outperforming the other methods by nearly 10%. This ECG person-identification modality would be a useful supplement for conventional biometrics, such as fingerprint and palm recognition systems.

Resolving the Limb Position Effect in Myoelectric Pattern Recognition

Reported studies on pattern recognition of electromyograms (EMG) for the control of prosthetic devices traditionally focus on classification accuracy of signals recorded in a laboratory. The difference between the constrained nature in which such data are often collected and the unpredictable nature of prosthetic use is an example of the semantic gap between research findings and a viable clinical implementation. In this paper, we demonstrate that the variations in limb position associated with normal use can have a substantial impact on the robustness of EMG pattern recognition, as illustrated by an in- crease in average classification error from 3.8% to 18%. We propose to solve this problem by: 1) collecting EMG data and training the classifier in multiple limb positions and by 2) measuring the limb position with accelerometers. Applying these two methods to data from ten normally limbed subjects, we reduce the average classification error from 18% to 5.7% and 5.0%, respectively. Our study shows how sensor fusion (using EMG and accelerometers) may be an efficient method to mitigate the effect of limb position and improve classification accuracy.

Examining the adverse effects of limb position on pattern recognition based myoelectric control

Pattern recognition of myoelectric signals for the control of prosthetic devices has been widely reported and debated. A large portion of the literature focuses on offline classification accuracy of pre-recorded signals. Historically, however, there has been a semantic gap between research findings and a clinically viable implementation. Recently, renewed focus on prosthetics research has pushed the field to provide more clinically relevant outcomes. One way to work towards this goal is to examine the differences between research and clinical results. The constrained nature in which offline training and test data is often collected compared to the dynamic nature of prosthetic use is just one example. In this work, we demonstrate that variations in limb position after training can have a substantial impact on the robustness of myoelectric pattern recognition.

A Physical Action Potential Generator : Design, Implementation and Evaluation

The objective was to develop a physical action potential generator (Paxon) with the ability to generate a stable, repeatable, programmable, and physiological-like action potential. The Paxon has an equivalent of 40 nodes of Ranvier that were mimicked using resin embedded gold wires (Ø = 20 μm). These nodes were software controlled and the action potentials were initiated by a start trigger. Clinically used Ag-AgCl electrodes were coupled to the Paxon for functional testing. The Paxons action potential parameters were tunable using a second order mathematical equation to generate physiologically relevant output, which was accomplished by varying the number of nodes involved (1–40 in incremental steps of 1) and the node drive potential (0–2.8 V in 0.7 mV steps), while keeping a fixed inter-nodal timing and test electrode configuration. A system noise floor of 0.07 ± 0.01 μV was calculated over 50 runs. A differential test electrode recorded a peak positive amplitude of 1.5 ± 0.05 mV (gain of 40x) at time 196.4 ± 0.06 ms, including a post trigger delay. The Paxons programmable action potential like signal has the possibility to be used as a validation test platform for medical surface electrodes and their attached systems.

Hidden Markov model classification of myoelectric signals in speech

It has been demonstrated that myoelectric signal (MES) automatic speech recognition (ASR) using an hidden Markov model (HMM) classifier is resilient to temporal variance, which offers improved robustness compared to the linear discriminant analysis (LDA) classifier. The overall performance of the MES ASR can be further enhanced by optimizing the features and structure of the HMM classifier to improve classification rate. Nevertheless, the HMM classifier has already shown that it would effectively complement an acoustic classifier in a multimodal ASR system.

Myo-electric signals to augment speech recognition

It is proposed that myo-electric signals can be used to augment conventional speech-recognition systems to improve their performance under acoustically noisy conditions (e.g. in an aircraft cockpit). A preliminary study is performed to ascertain the presence of speech information within myo-electric signals from facial muscles. Five surface myo-electric signals are recorded during speech, using Ag−AgCl button electrodes embedded in a pilot oxygen mask. An acoustic channel is also recorded to enable segmentation of the recorded myo-electric signal. These segments are processed off-line, using a wavelet transform feature set, and classified with linear discriminant analysis. Two experiments are performed, using a ten-word vocabulary consisting of the numbers ‘zero’ to ‘nine’. Five subjects are tested in the first experiment, where the vocabulary is not randomised. Subjects repeat each word continuously for 1 min; classification errors range from 0.0% to 6.1%. Two of the subjects perform the second experiment, saying words from the vocabulary randomly; classification errors are 2.7% and 10.4%. The results demonstrate that there is excellent potential for using surface myo-electric signals to enhance the performance of a conventional speech-recognition system.

A multi-modal approach for hand motion classification using surface EMG and accelerometers

For decades, electromyography (EMG) has been used for diagnostics, upper-limb prosthesis control, and recently even for more general human-machine interfaces. Current commercial upper limb prostheses usually have only two electrode sites due to cost and space limitations, while researchers often experiment with multiple sites. Micro-machined inertial sensors are gaining popularity in many commercial and research applications where knowledge of the postures and movements of the body is desired. In the present study, we have investigated whether accelerometers, which are relatively cheap, small, robust to noise, and easily integrated in a prosthetic socket; can reduce the need for adding more electrode sites to the prosthesis control system. This was done by adding accelerometers to a multifunction system and also to a simplified system more similar to current commercially available prosthesis controllers, and assessing the resulting changes in classification accuracy. The accelerometer does not provide information on muscle force like EMG electrodes, but the results show that it provides useful supplementary information. Specifically, if one wants to improve a two-site EMG system, one should add an accelerometer affixed to the forearm rather than a third electrode.

Security-Monitoring using Microphone Arrays and Audio Classification

This paper proposes a security-monitoring instrument that can detect and classify the location and nature of different sounds in a room. The instrument is reliable and robust even in the presence of reverberation and in low signal to noise ratio conditions. This paper proposes a new algorithm for classifying first an audio segment as speech or nonspeech then classifies the nonspeech audio segment into its own audio type. The algorithm divides an audio segment into frames, estimates the presence of pitch in each frame, and calculates a pitch ratio parameter. This parameter is then used to classify the audio segment. The threshold used in calculating this parameter is adapted to accommodate different environments. Nonspeech audio segment has further classification using time delayed neural network to be classified into it is own type. The performance of the proposed algorithm is evaluated for different signal-to-noise ratios using a library of audio segments. The library includes speech segments and nonspeech segments such as windows breaking and footsteps. Using 0.4 second segments it is shown that the proposed algorithm can achieve an average correct decision for 94.5% of the reverberant library and 95.1% of the nonreverberant library