Several Key Technologies In Surface Electromyography Signal Detection & Processing

Surface electromyographic signal (sEMG) is the electrophysiological signal concomitant with musculations. There're two research focuses about sEMG:Human action recognition based on sEMG and sEMG decomposition. The former aims for providing a more natural, more convenient and more effective way of computer human interaction (CHI), which could be widely applied in artificial limb controlling, mobile device handling, gesture language recognition, sport electronics, games/ entertainments, tactical commanding, etc. The latter aims for extracting information about motor unit recruitment/firing and MUAP waveforms, so as to studying the control mechanism of neuromuscular system, and providing evidences for diagnosing neuromuscular diseases.Compared with intramuscular EMG (iEMG) signal, the sEMG signal detection can be widely applied for the advantage of non-invasive. However, for detected from skin surface with bigger sized electrodes, and influenced by low-pass-filter effect of muscle, hypoderm and skin, the sEMG signal has a lower signal noise ratio, a higher waveform dissimilation, and a higher difficulty in detections and processes. As a result, in the fields of sEMG decomposition, and human action recognition/CHI based on sEMG, there're still several issues of technologies to solve, for example: Designs of sEMG electrodes in special applications; The noises, impedances and adaptabilities of electrodes; Ways of reducing power frequency interferences; Designs of low noise, low power and wearable sEMG signal acquisition systems for CHI; High efficiency, real time and structurized method for action recognition; Recognizing method for continuous gesture actions; High density, high performance and low noise electrodes array for acquiring temporal and spacial information from sEMG; etc. Focus on these issues, the main work and achievement of the dissertation could be presented as follows:1. Designs and analysis of high performance electrodes for sEMG detection. Revolving closely around the applications of sEMG signal, many kinds of convenient and durable dry electrodes were developed, including line-shaped electrodes which are simple, convenient and durable; printed electrodes which have devisable shapes and can be comfortably wore; spring probe electrodes which have good temporal/ spacial resolutions and can be well contacted. The noises of these electrodes were tested and analyzed by experiments; The impedance characteristics of these electrodes including time-varying behavior, frequency response, effect of pressure and skin pretreatment, were studied using home-developed vector admittance instrument; Motion artifacts of the electrodes were filtered out by improving the instrumentation amplifier circuits. All these methods and results could provide universal meanings for works of sEMG signal detection.2. The power frequency interference reduction (PFIR) in the weak signal detection of sEMG. Compared with traditional PFIR algorithms, the time-domain grouped filtering based on PLL synchronized clock which presented as a novel software-hardware combined PFIR method, has the advantage of fast, efficient and structurized. A novel isolated power supply with the structure of double transformers was developed for lower input/output noises and lower common mode interferences, capitalizing on the effects of complemental counteraction. In addition, metal plate shielding and battery supplied acquisition also achieved satisfying PFIR effects.3. Designs of sEMG signal acquisition systems (SAS) for multiple applications. Several kinds of sEMG signal acquisition systems with high signal noise ratios and high common mode rejection ratios were developed, including sEMG SAS with active electrodes, FM wireless sEMG SAS,2.4GHz wireless digital sEMG/ Acceleration SAS and Bluetooth sEMG/Acceleration SAS. With improved performances, miniaturized sizes, lightened weights, reduced power dissipations, enhanced portabilities and wearabilities by multiple ways and techniques, the sensors and systems are able to provide several novel ways of interaction for wearable computers and mobile hand-held devices.4. Real-time gesture recognition based on sEMG and its realization on FPGA. For the demands of miniaturized and integrated interaction devices, some reliable, fast and structurized algorithms were used for real-time gesture recognition, and realized on an FPGA platform.3 stage cache controlling method, fast action detection method based on looped queue, feature extraction method based on mean absolute value and AR coefficients, classifiers based on weighted Euclidean distance or Bayes method (LDC) were used in the real-time gesture recognition. A platform for real-time gesture recognition was also built, by which the recognition speeds, success rates and user adaptabilities of the methods were tested and analyzed.5. Continuous gesture recognition based on sEMG. Based on the feature of short time energy, a method of gesture recognition that judges by signal energy's magnitude and similarity to patterns was presented for continuous gesture recognition. In this way, recognition is immediately proceeded within each data sample duration, and the continuous recognition is independent of "action segment"6. Research on high density electrodes for acquiring temporal and spacial information of sEMG. Double differential electrodes (DDE) were introduced in studies about sEMG of hand gestures. It was proved in the experiments that the positional resolution was improved by the DDE, at the same time, the DDE was more sensitive to positional deviations. A kind of integrated sEMG electrodes array sensor with a first stage amplifier was developed, which managed to reduce interferences and noises by the integrated way. Several studies was carried out with the sensor:The measurement of muscle fiber conduction velocity (MFCV); The MUAP durations of printed electrodes and spring probe electrodes, approximate evaluated by the main peak width of signal's autocorrelation; The improvement of MUAP resolution by spacial filtering.The research is supported by the National "863" Project "Research on EMG sensors and accelerometers based hand gesture interaction devices" (2009AA01Z322) and the National Natural Science Foundation of China "Surface EMG decomposition based on linear varied force and temporal/spacial information from multiple channels" (30870656).