D. F. Lovely

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.

Steady-state and dynamic myoelectric signal compression using embedded zero-tree wavelets

Within the field on biomedical engineering, the majority of compression research has focused on encoding medical images, electrocardiograms, and electroencephalograms. Although long-term myoelectric signal (MES) acquisition is important for neuromuscular system analysis and telemedicine applications, very few studies have been published on MES compression. This research investigates static and dynamic MES compression using the embedded zerotree wavelet (EZW) compression algorithm and compares its performance to a standard wavelet compression technique.

Real-time compression of myoelectric data utilising adaptive differential pulse code modulation.

The myoelectric signal, obtained by either surface or needle electrodes, is used in many areas of clinical research and diagnosis. The conventional method of storing such information is in digitised form on a computer. However, the bandwidth of the signal and the required resolution result in large memory requirements. Adaptive differential pulse code modulation is investigated as a method of reducing the memory requirements for myoelectric data storage. In this scheme, a 12-bit sample is reduced to four bits, thus reducing the memory requirements by a factor of three. In reality, this compression ratio is closer to 4∶1 owing to the fact that the widths of most memories are organised as multiples of eight bits.

Multiexpert automatic speech recognition using acoustic and myoelectric signals

Classification accuracy of conventional automatic speech recognition (ASR) systems can decrease dramatically under acoustically noisy conditions. To improve classification accuracy and increase system robustness a multiexpert ASR system is implemented. In this system, acoustic speech information is supplemented with information from facial myoelectric signals (MES). A new method of combining experts, known as the plausibility method, is employed to combine an acoustic ASR expert and a MES ASR expert. The plausibility method of combining multiple experts, which is based on the mathematical framework of evidence theory, is compared to the Borda count and score-based methods of combination. Acoustic and facial MES data were collected from 5 subjects, using a 10-word vocabulary across an 18-dB range of acoustic noise. As expected the performance of an acoustic expert decreases with increasing acoustic noise; classification accuracies of the acoustic ASR expert are as low as 11.5%. The effect of noise is significantly reduced with the addition of the MES ASR expert. Classification accuracies remain above 78.8% across the 18-dB range of acoustic noise, when the plausibility method is used to combine the opinions of multiple experts. In addition, the plausibility method produced classification accuracies higher than any individual expert at all noise levels, as well as the highest classification accuracies, except at the 9-dB noise level. Using the Borda count and score-based multiexpert systems, classification accuracies are improved relative to the acoustic ASR expert but are as low as 51.5% and 59.5%, respectively.

A multi-expert speech recognition system using acoustic and myoelectric signals

Performance of conventional automatic speech recognition systems, which uses only the acoustic signal, is severely degraded by acoustic noise. The myoelectric signal from articulatory muscles of the face is proposed as a secondary source of speech information to enhance conventional automatic speech recognition systems. An acoustic speech expert and myoelectric speech expert are combined using a novel approach based on evidence theory. Data were collected from 5 subjects across an 18 dB range of noise levels. The classification rate of the acoustic expert decreased dramatically with noise, while the myoelectric signal expert remained relatively unaffected by the noise. This method of combining experts is able to dynamically track the reliability of experts. Classification rates of the multi-expert system were better or near either individual expert at all noise levels.

Identification of noise sources in surface recording of spinal somatosensory evoked potentials

The poor signal-to-noise ratio associated with the acquision of evoked potentials is a well established fact. The problem is compounded if non-invasive techniques, using surface electrodes, are employed. The paper identifies several sources of noise associated with the acquisition of spinal somatosensory evoked potentials using surface electrodes. In addition, the relative contribution of these sources is determined experimentally for six spinal levels ranging from lower lumbar to upper thoracic. These data will prove useful in the design of digital signal processing schemes such as adaptive noise cancellation, where levels of uncorrelated noise severely limit system performance.

A Bus-Based Smart Myoelectric Electrode/Amplifier—System Requirements

The evaluation of a bus-based smart myoelectric electrode/amplifier is described that is to be used in conjunction with a multi-function prosthetic hand controller. The smart electrode/amplifier was designed to meet power consumption and size specifications of commercially available myoelectric amplifiers used for prosthetic control applications while providing a number of additional features. This paper investigates the electrode/amplifier requirements for a pattern classifier system and compares them to those currently accepted within a clinical setting. System testing and evaluation was performed with both normally limbed subjects and traumatic amputees.

Reduction of evoked potential measurement time by a TMS320 based adaptive matched filter

In some clinical centres, somatosensory evoked potentials are used for the assessment of neurological function during surgical procedures on the spine. As these potentials are heavily contaminated in background noise, ensemble averaging coupled with invasive instrumentation is routinely used to enhance the signal. However, this procedure is very time consuming, often taking several minutes. In this paper, an adaptive matched filter has been used to dramatically reduce this measurement time to around 20 seconds, even when employing non-invasive surface electrodes. This filter has been implemented in real-time by using a TMS320C25 digital signal processor and results are presented for signal acquisition both at the L1-T12 and T5-T6 spinal levels. In addition, the adaptive nature of this filter allows the tracking of slowly changing parameters within the evoked potential with time.

Myoelectric signal compression using zero-trees of wavelet coefficients

Recent progress in the diagnostic use of the myoelectric signal for neuromuscular diseases, coupled with increasing interests in telemedicine applications, mandate the need for an effective compression technique. The efficacy of the embedded zero-tree wavelet compression algorithm is examined with respect to some important analysis parameters (the length of the analysis segment and wavelet type) and measurement conditions (muscle type and contraction type). It is shown that compression performance improves with segment length, and that good choices of wavelet type include the Meyer wavelet and the fifth order biorthogonal wavelet. The effects of different muscle sites and contraction types on compression performance are less conclusive.A comparison of a number of lossy compression techniques has revealed that the EZW algorithm exhibits superior performance to a hard thresholding wavelet approach, but falls short of adaptive differential pulse code modulation. The bit prioritization capability of the EZW algorithm allows one to specify the compression factor online, making it an appealing technique for streaming data applications, as often encountered in telemedicine.