Design of Context-Aware Exercise Measurement SoC Based on Electromyogram and Electrocardiogram Seongsoo Lee Abstract Exercise measurement is an important application of smart healthcare devices. Conventional approaches cannot measure indoor exercise or resistive exercise. In this paper, we propose a context-aware exercise measurement system-on-chip (SoC) based on electromyogram (EMG) and electrocardiogram (ECG). It consists of 3 EMG sensors, analog circuits for signal acquisition and digital conversion, and digital circuits for signal correction and exercise calculation. Keywords Context-Aware, Electrocardiogram, Electromyogram, Exercise Measurement, System-on-Chip E I. INTRODUCTION XERCISE measurement is an important application of smart healthcare devices. Digital exercise assistant is one of the most fast-growing markets in digital equipment. Many approaches are proposed in exercise measurement, but most conventional approaches suffer from several severe problems. Electromyogram (EMG), micro-scale electrical signals from muscles, contains important information on the movement of body. Similarly, electrocardiogram (ECG), micro-scale electrical signals from heart, also contains important information on the body condition. Therefore, accurate exercise amount with less measurement restriction is possible by observing electromyogram and electrocardiogram. In this paper, we propose a context-aware exercise measurement system-on-chip (SoC) based on EMG and ECG measurement. It consists of 3 EMG sensors, analog circuits for signal acquisition and digital conversion, and digital circuits for signal correction and exercise calculation. In conventional approaches, separate ECG sensors are needed in ECG measurement. However, the proposed SoC extracts ECG waveform from EMG waveform, so it does not require ECG sensors. Furthermore, the proposed SoC can measure both aerobic exercise and resistive exercise, while conventional approaches can measure only aerobic exercise. For convenient and comprehensive diet planning, it is advantageous to convert exercise to calorie. Exercise-calorie Manuscript received July 28, 2011. This work was sponsored by ETRI System Semiconductor Industry Promotion Center, Human Resource Development Project for SoC Convergence. Seongsoo Lee is with the School of Electronic Engineering, Soongsil University, Seoul, 156-743, Korea (e-mail: sslee@ssu.ac.kr) conversion requires basal metabolic rate (BMR) based on weight, height, and age. The proposed SoC also provides exercise-calorie conversion based on BMR. II. CONVENTIONAL METHODS OF EXERCISE MEASUREMENT Conventional methods of exercise measurement are classified into three categories, i.e. global positioning system (GPS)-based method, piezo-electric ceramic (PEC)-based method, and accelerometer-based method. A. GPS-Based Method Sprint proposed BIMAactive for outdoor exercise measurement, as shown in Fig. 1. BIMActive measures outdoor exercise by measuring running time, distance and speed via GPS. It also provides online uploading of exercise data. It also provides online audiovisual exercise tutorial for beginners. MagellanGPS proposed Triton for outdoor exercise measurement, as shown in Fig. 2. Triton shows exercise amount, running time, distance and speed, and calorie consumption based on GPS measurement. GPS-based method can perform precise measurement of outdoor exercise. However, it cannot measure indoor exercise due to GPS signal coverage. Furthermore, it cannot measure resistive exercise such as pull-up bar and weight training, since there is no position displacement during resistive exercise. B. PEC-Based Method Nike and Apple proposed Sports ipod for running exercise measurement by embedding piezo-electric sensors at the bottom of Nike running shoes, as shown in Fig. 3. It measures the number of walking/running steps, and transmits this information to ipod, where running time distance and speed, and calorie consumption is calculated and displayed. Similarly, Adidas proposed One Shoes based on piezo-electric sensors, as shown in Fig. 4. Unlike GPS-based method, PEC-based method can measure indoor exercise such as treadmill. However, it only counts the number of walking/running steps, so the exercise measurement is quite inaccurate. Like GPS-based method, PEC-based method cannot measure resistive exercise such as pull-up bar and weight training, since there are no walking/running steps during resistive exercise. ISBN: 978-1-61804-035-0 105
Fig. 1 BIMActive Fig. 5 Actiwatch Score Fig. 2 Triton Fig. 6 Active Style Pro Fig. 3 Sports ipod Fig. 4 One Shoes C. Accelerometer-Based Method BMedical proposed Actiwatch Score for indoor/outdoor exercise measurement based on accelerometer, as shown in Fig. 5. Similarly, Omron Healthcare proposed Active Style Pro, as shown in Fig. 6. Accelerometer-based method is most frequently used in exercise measurement. It measures the acceleration of vertical and horizontal movement of the body. By integrating the acceleration along time, exercise amount and its corresponding calorie consumption are calculated. Accelerometer-based method is quite accurate in normal exercise conditions. However, it cannot calculate precisely in fast or slow exercise. Similarly, it overestimates exercise amount when running up along stairs or slope. Like GPS-based method and PEC-based method, accelerometer-based method cannot measure resistive exercise such as pull-up bar and weight training, since there is no motion with acceleration during resistive exercise. III. EXERCISE MEASUREMENT USING ELECTROMYOGRAM Electromyogram (EMG) is micro-scale electrical signals from muscles. It has wide frequency range from 0.1 ~ 500 Hz. Originally, a needle is inserted into muscles to measure EMG signal (nemg), but recently an electrode is attached to skin ISBN: 978-1-61804-035-0 106
Fig. 7 semg sensors for the proposed SoC Fig. 8 Extraction of 3 different frequency ranges Fig. 9 ECG signal extraction Fig. 10 High-frequency and low-frequency EMG signal extraction surface (semg). In the proposed SoC, 3 semg sensors are attached in belts, as shown in Fig. 7. The proposed SoC measures semg signals and extracts 3 different frequency ranges, i.e. low-frequency EMG signal, high-frequency EMG signal, and ECG signal, as shown in Fig. 8. ECG signal extraction is illustrated in Fig. 9. First, signal in ECG frequency range is extracted by bandpass filter (1-50 Hz). Then, baseline value for absolute amplitude value algorithm [1] is determined. Then, absolute amplitude values are extracted using the baseline value. Then, threshold value for R-wave peak point detection is determined. Then, R-ware peak point is detected using the threshold value. Finally, heartbeat rate is calculated by counting R-wave peak point [2]. High-frequency and low-frequency EMG signals are extracted as illustrated in Fig. 10. High-frequency EMG signal represents resistive exercise and arm/leg movement [3]. Highpass filter (> 30 Hz) is used for high-frequency EMG signal extraction. Low-frequency EMG signal represents breath and body movement [3]. Lowpass filter (< 1 Hz) is used for low-frequency EMG signal extraction. After exercise amount is calculated, it should be converted in calorie consumption. Exercise-calorie conversion requires basal metabolic rate (BMR) based on weight, height, and age. In the proposed SoC, Harris-Benedict equation (B.E.E) method [4] and Institute of Medicine of the National Academies (I.M.N.A) method [5] are used as follows. Harris-Benedict Equation (B.E.E) Method - Men: 66.47+(13.75 weight in kilos)+(5.0 height in cm)- (6.76 age in year) - Women: 655.1+(9.56 weight in kilos)+(1.85 height in cm) -(4.68 age in year) Institute of Medicine of the National Academies (I.M.N.A) Method - Men: 293-(3.8 age in year)+(456.4 height in meter)+ (10.12 weight in kilos) - Women: 247-(2.67 age in year)+(401.5 height in meter)+ (8.60 weight in kilos) IV. SOC IMPLEMENTATION Architecture and design target of the proposed context-aware exercise measurement SoC is illustrated in Fig. 11. It consists of (1) 3 semg sensors, (2) analog circuits for signal acquisition and digital conversion, and (3) digital circuits for signal correction and exercise calculation. Architecture of signal acquisition circuit is illustrated in Fig. 12. Digital circuit consists of digital filters, 8051 microprocessor, and universal I/Os. 2 types of digital filters are used, i.e. finite impulse response (FIR) filter and infinite impulse response (IIR) filter. To reduce chip area, serial processing architectures are used, as illustrated in Fig. 13 and Fig. 14. Universal I/O is an integrated block of various I/Os such as UART, SPI, and I2C, as illustrated in Fig. 15. The proposed SoC has its own data transfer protocol between SoC and external PC/equipment, as shown in Fig. 16. Fig. 17 shows the post simulation waveforms of the proposed SoC. ISBN: 978-1-61804-035-0 107
Recent Researches in Circuits, Systems, Control and Signals Fig. 11 Architecture and design target of the proposed context-aware exercise measurement SoC Vin LPF (<1Hz) ADC1 D out1 BPF (1~60Hz) ADC2 Dout2 HPF (>30Hz) ADC3 D out3 LNA Fig. 12 Architecture of signal acquisition circuit Fig. 16 Data transfer protocol Fig. 13 Serial processing architecture of FIR filter Fig. 14 Serial processing architecture of IIR filter Fig. 17 Post simulation waveforms of the proposed SoC Fig. 15 Architecture of universal I/O ISBN: 978-1-61804-035-0 108
V. CONCLUSION A context-aware exercise measurement SoC based on electromyogram and electrocardiogram is proposed. It consists of 3 EMG sensors, analog circuits for signal acquisition and digital conversion, and digital circuits for signal correction and exercise calculation. The proposed SoC is under design and will be fabricated in early 2012. REFERENCES [1] C. Min and T. Kim, Development of Signal Detection Methods for ECG (Electrocardiogram) based u-healthcare Systems, Journal of IEEK, vol. 46, no. 6, pp. 641-649, 2009. [2] C. Min and T. Kim, ECG Based Patient Recognition Model for Smart Healthcare Systems, Lecture Notes in Computer Science, vol. 3398, pp.159-166, 2005. [3] H. Kim, C. Min and T. Kim, Adaptable Noise Reduction of ECG Signals for Feature Extraction, Lecture Notes in Computer Science, vol. 3973, pp. 586-591, 2006. [4] J. Harris and F. Benedict, A Biometric Study of Human Basal Metabolism, Proceedings of National Academy of Science, 1918. [5] Institute of Medicine of the National Academies, Dietry Reference Intakes for Energy, Carbohydrate, Fiber, Fat, Fatty Acids, Cholesterol, Protein, and Amino Acids, The National Academies Press, 2009. ISBN: 978-1-61804-035-0 109