A Novel Approach for Simulation, Measurement and Representation of Surface EMG (semg) Signals

Transcription

1 A Novel Approach for Simulation, Measurement and epresentation of Surface EMG (semg) Signals Anvith Katte Mahabalagiri, Khadeer Ahmed, Fred Schlereth Syracuse University, Syracuse, NY USA Abstract- In this paper, we describe new methods for the simulation, measurement and representation of semg signals. With regard to simulation, we choose a 2-D state space model and suggest a 3 D model which can account for in- homogeneities, nonlinearities and memory in the medium and which can be hardware accelerated through FPGA. With regard to measurement we use surface electrodes with a new amplifier circuit topology, which mitigates the effects of pickup and artifacts. With regard to representation, we describe a method for using wavelets which shows promise for isolating signals of interest. 0 v 1 Muscle Fat and Tissue Skin V 0 V 3 S INTODUTION Many biophysical questions that might be answered by observing SEMG signals remain unanswered without proper knowledge of volume conduction and its implementation in models. Several approaches have been proposed in this regard [1], [3], [4], [5]. A 2-D model and its 3-D representation based on a 3 layer medium model of muscle fat and skin [6] is shown in Figure 1 and 2 respectively, which can account for the major phenomena affecting the signals generated by the muscles as they pass from the muscle to the skin. These include inhomogeneities, nonlinearities and memory in the medium. Furthermore we show how to formulate a discrete representation of the continuous model which can be implemented on an FPGA, making use of the parallel computing resources of the FPGA. The advantage is greatly improved speed of execution as compared with implementation on a workstation. Our desire is to combine a volume model, a new measurement topology and a wavelet based representation to use the traveling wave nature of Motor Unit Action Potential (MUAPs). The goal is to detect signal patterns which can be used in clinical settings such as handwriting analysis, as an aid to detecting early onset of conditions such as Alzheimer's disease. A longer term goal is to be able to detect finger movement with sufficient accuracy and sensitivity to provide a means for cursor control on a computer. Applications such as these require a common methodology for simulation, measurement and representation semg signals. In this paper, we propose such a methodology. Section I describes a method to build a reliable simulation model by comparing the measured values with the model and finding optimal positions 0 v 2 0 v 3 V 1 V 2 Fig. 1. Volume conduction model for the placements of sensor arays for detection of given MUAP. Sections II and III describe a method of measurement and representation based on the simulation model. I STATE SPAE MODEL In this paper, we consider a 3 layered model [6]. The muscle fiber works due to motion of ions, i.e. it operates as a current source. In our model, we have represented the same current source as a voltage source for developing the state space model. The capacitance across the source is required due to the nature of muscle fiber. The fat and tissue layer is approximated as a low pass filter [7]. Finally the skin layer is represented as simple resistance. The values used for the model are as suggested by [6], [7]. The electrical circuit for this model is as shown in figure 1. For simplicity sake, we have considered same values of 0, and for all the zones however the model can be built with different values in a V 4 V 5 S S /11/$ IEEE 476 Asilomar 2011

2 similar way. The voltages represented in the circuit are the voltages across the capacitors, not with respect to ground. The above 2-D model can be easily extended to a 3-D state space model as shown in figure 2. Here the MUAPs are current sources. Storage in the medium is due to the capacitors at each of the internal nodes. A resistive interconnection matrix is assumed to account for transmission of the signal from the source to the surface. A third dimension is easily added by layering the above diagram into the paper and connecting the layers through resistors. MUAP along the fiber Muscle Fibre to Surface Fig. 2. Volume conduction model The state space form for the circuit in figure 1 is given by (1) Where, is a vector containing the first derivatives of the state variables, is the vector of input sources and A and B are the co-efficient matrices. Figure 3 shows the form of the state space matrix for the model (2) We perform discretization using equation (2) to model the above system in Simulink. Figure 4 shows a Simulink implementation for the above circuit. This implementation has many multiplier-accumulator blocks. Given the Simulink implementation, it is a simple matter to program the FPGA. This model is very general and can account for a myriad of real world transmission and signal generation effects and because of the FPGA implementation, it can have high resolution with very little impact on computation time compared to workstation implementation. In future, if physical evidence suggests, it would be a simple modification to include inductance as well. We hope that this simulation model will serve to inspire biophysicists to develop more detailed physical models than are presently known. Through the use of such a model it will be possible to perform parametric studies that would take an inordinate amount of time if performed with direct measurements. II SIGNAL MEASUEMENT The major factors affecting the measurement of MUAPs with surface electrodes are 60 Hz pickup, motion artifacts and the fact that each sensor is responding to the summation of a large number of individual signals. The amplifiers and subsequent A/D converters have the task of producing a high fidelity reproduction of the physical signals at the skin surface. We suggest a circuit topology that takes into account of variations in impedance between sensor and skin and provides better reduction of motion artifacts, as compared with conventional methods. Fig. 3. State Space Matrix 477

3 Fig. 4. Simulink Model The circuit, shown in Figure 5 consists of a D coupled preamp, diff amp, and post amp, feeding an A/D converter. Additional filtering, if needed, is provided by the digital signal processing stage. In this circuit we use three sensors, with one common sensor among the pre amps to form one channel of measurement. Pre Amp Pre Amp Diff Amp Post Amp A/D Figure 5 Amplifier System III SIGNAL EPESENTATION Wavelets are key to the representation of semg signals because 1) their ability to portray a 3-D map of time and frequency 2) their ability to detect particular wave shapes buried in background noise and interfering signals. Most important is that MUAPs are known to travel in waves along the muscle fibers and wavelets in conjunction with a spatial array of sensors offer enough structure to derive information from these traveling signals using only the surface electrodes. Once the MUAPS are represented as wavelets then their processing can be done parallel and efficiently on an FPGA. This enables real time detection and processing of large number of channels compared to workstation implementation. 478

4 Fig. 6. Simulink model output- V1, V2 and V3 represent the MUAP at three different time frames. Vop1, Vop2, Vop3 are the outputs across resistance s IV SIMULATION ESULTS Figure 6 shows the output response of the Simulink model in figure 4, based on the state space matrix in figure 3 for the circuit model in figure 1. Input is a sample signal of the brachial biceps from the dataset of the EMG labs [2]. The input signal is through three different sources at three different time frames and their responses at respective positions on the skin surface representing the traveling nature of the MUAP. We can observe that V1 has the maximum effect on output Vop1 and similarly for the other sources. We can also observe the low pass effect of the medium on the source signal which is similar to real world measurements. For the sake of clarity we have exaggerated the source delays. Though the effect of a source is small on neighboring zones, the effect is cumulative and becomes prominent when the delays are reduced as in real situations. The measured output in real situation would be the summation of several such above mentioned outputs based on sensor placements. 479

5 V ONLUSION AND FUTUE WOK In this paper, we have presented a model for simulation of semg signals using a state space based volume conduction model. This model helps in better understanding the conduction of signals to the surface through different layers of fat, muscle and skin. The results are not an exact simulation but rather a conceptual proof to suggest a new promising methodology for simulation. We have further suggested a circuit topology for measuring the signal and representing it using wavelets based on the above model. The simulation model together with the measurement and representation methodology can help us in efficient detection of the signal with minimal impact from noise and artifacts. We have built and tested hardware and simulation models and are currently performing experiments to determine the efficacy of this new approach. Work needs to be done in validating the simulated signals with actual measurements. These validated results will be used to determine optimal sensor placement for semg signals. EFEENES [1]. Electromyography, Physiology, Engineering and Noninvasive Applications, edited by. Merletti and P. Parker, H 8, 2004, IEEE. [2]. McGill K.[Online dataset 001, available at [3]. Lowery, M.M.; Stoykov, N.S.; Dewald, J.P.A.; Kuiken, T.A.; "Volume conduction in an anatomically based surface EMG model," Biomedical Engineering, IEEE Transactions on, vol.51, no.12, pp , Dec [4]. Gygi, A.E.; Moschytz, G.S.;, "Low-pass filter effect in the measurement of surface EMG," omputer-based Medical Systems, Proceedings, Tenth IEEE Symposium on, vol., no., pp , Jun 1997 [5]. K. oeleveld, J. H. Blok, D. F. Stegeman, A. van Oosterom, Volume conduction models for surface EMG; confrontation with measurements, Journal of Electromyography and Kinesiology, Volume 7, Issue 4, December 1997, Pages , ISSN , DOI: /S (97) [6]. Farina, D.; Merletti,.; "A novel approach for precise simulation of the EMG signal detected by surface electrodes," Biomedical Engineering, IEEE Transactions on, vol.48, no.6, pp , Jun 2001 [7]. Fansan Zhu; Levin, N.W.; "An electrical resistivity model of segmental body composition using bioimpedance analysis," Engineering in Medicine and Biology Society, Proceedings of the 25th Annual International onference of the IEEE, vol.3, no., pp Vol.3, Sept