EASTERN MEDITERRANEAN UNIVERSITY FACULTY OF ENGINEERING Electrical and Electronics Engineering Department

Size: px

Start display at page:

Download "EASTERN MEDITERRANEAN UNIVERSITY FACULTY OF ENGINEERING Electrical and Electronics Engineering Department"

Cornelia Joseph
6 years ago
Views:

1 EASTERN MEDITERRANEAN UNIVERSITY FACULTY OF ENGINEERING Electrical and Electronics Engineering Department Fall EEE 420 Project Report Ahmet Cem VARDAR Project Title: Heart Rate Monitor

2 TABLE OF CONTENTS TABLE OF CONTENTS... 2 INTRODUCTION... 3 CHAPTER I HARDWARE... 4 CHAPTER II SOFTWARE...6 SECTION 1) A/D CONVERSION TECHNIQUE:... 6 SECTION 2) SIMPLE SOLUTION: CHAPTER III - CONCLUSIONS

3 Introduction My main aim was to acquire cardio signals and measure the heart beat rate using a microcontroller. There exist various ways of observing the rythmia of heart. Main resources of observation are electrocardiographical signals, sound waves, and blood pressure. Several circuits designed for electrocardiography (ECG) amplification, and filtration. Since cardiac signals are carried by body fluids it also interferes with some other voltage sources in the body, and those voltages accounted as noise. So excellent amplification and filtration is required in this technique. Additionally use of such a technique may result in fatal injury or lost of life if proper isolation is not used. In this report I will discuss optical approach that benefits from blood pressure. This technique simplifies the circuitry and increases the safety. It enables us to get the heart beat rate and some other data (such as cholesterol level), but the acquired signal does not have those complex components which are necessary for cardiologic diagnoses. Discussions are divided into two subsections, first one is dedicated for the hardware part, and the second one is related to the software. 3

4 CHAPTER I HARDWARE Hardware part has three main elements, 1. Optical room, 2. Amplifier 3. Microcontroller The optical room consists of a high power light emitting diode and a light sensitive resistor (see FIG 1.1). Light sensitive resistor is a part which has varying resistance (10 ohm to 5 M ohm), dependent on the illumination. My optical room is designed such that no light waves could enter from any side when the finger is inside. FIGURE 1.1 Optical architecture. Amplifier is used to condition the voltage seen across the light sensitive resistor. In my circuit I have used OPA 2604 from Texas Instruments. I have chosen this IC because of its low drift characteristics. Additionally it has dual operational amplifier, so one of them could be used for trimming the input signal. Finally microcontroller is used to count the heart beats and show the rate on the display. At this point I have applied two approaches. One of them is using an analog to digital converter, filter the signal, and measure the frequency (hard way) and another way is directly trimming the heart input signal at the op-amp and send it in TTL form to one of the logical inputs of a simple microcontroller. 4

5 FIGURE 1.2 Hardware assembly phase. Test results showed that trimming the signal and acquiring form IO port is more reliable. Blood pressure variations were displayed on the color oscilloscope. A screen shot can be seen in FIG 1.3. FIGURE 1.3 Oscillospe screenshot. 5

CHAPTER II SOFTWARE Section 1) A/D Conversion Technique: For the first approach that utilizes the analog digital conversion, microcontrollers accumulator and memory limitations did not let me to

6 CHAPTER II SOFTWARE Section 1) A/D Conversion Technique: For the first approach that utilizes the analog digital conversion, microcontrollers accumulator and memory limitations did not let me to apply those calculations on board, but instead I have transferred the data serially via RS-232 serial communication at Bps. A simple Visual Basic programming was necessary to apply the DSP, plot the signal and show DSP results. FIGURE 2.1 Pulse signal after analog preprocessing and digitization. 6

7 As seen in the above figure (see FIG 2.1) user has some settings, which also enables to see the original signal or the differentiated one. To determine the pulse rate the PC software first differentiates the pulse signal ( p[ i ] ). Numerical differentiation means to calculate the difference between two samples. So the differentiated signal is: pd [ i ] = p[ i ] - p[ i-dt ] The distance dt can be selected by the user and has an additional "filter effect". Small distances dt will enhance high frequency fluctuations and greater distances will enhance low frequency signals. FUGURE 2.2 Differentiated pulse signal of last figure (left) and corresponding autocorrelation function (right). 7

From the differentiated pulse signal the autocorrelation function is calculated. The strength of the function is that it is able to detect periodic parts hidden in a very noisy signal.

8 From the differentiated pulse signal the autocorrelation function is calculated. The strength of the function is that it is able to detect periodic parts hidden in a very noisy signal. You can see the effectiveness of autocorrelation function in the following two figures (see FIG 2.2 and 2.3). FIGURE 2.2 Pulse signal disturbed by high frequency noise (left) and corresponding autocorrelation function (right). 8

9 FIGURE 2.3 Pulse signal disturbed by high and low frequency noise (left) and corresponding autocorrelation function (right). Originally the autocorrelation function is defined for an infinite number of samples. It s obvious that an infinite measurement time is not possible and we want to monitor changes in the heart rate, so we can only determine the autocorrelation over a limited time window of n samples. To solve this problem a moving average of the correlation factors between samples with different time distances τ is calculated. (Windowing principle as discussed in Texas Instruments sbao0092.pdf application note Incorporating MSC1210 into electronic weight scale systems). In the above equation pd[ i ] is the most recent value of the differentiated pulse signal. The time window has a length of n samples. The older values are read form a circular buffer. (The buffer size must be n + τmax; a value of 300 is used for τmax) 9

10 The computational effort to calculate the ACF is significantly reduced using a recursive algorithm: This algorithm allows the ACF to be updated for every new value received by the PC. The ACF display is redrawn when all values read form the receive buffer are processed. This result in live display (tested at a 2.8 GHz Pentium processor). (The length of the time window for the moving average can be set by the user as "buffer size". The "reset" button sets the average buffer to zero.) Having calculated the autocorrelation function it still remains to determine the pulse rate. This is done by searching the highest peak in the ACF. Because normally at time zero the ACF has its absolute maximum only the region after the first zero crossing is examined. The time coordinate τ1 of the highest peak is taken as "first approach" for the pulse period. However especially "good" pulse signals will produce several maxima of almost the same height in their ACF. The true pulse period could therefore be an integer fraction of the period time τ1 found as "first approach". Therefore the ACF is examined at time values which are integer fractions of τ1. The lowest of this time values at which the ACF exceeds 80% of the absolute maximum found before is defined as the "true" signal period. (80% was empirically or approximately found to be appropriate in most cases.) From the signal period the pulse rate in counts per second is calculated and displayed in the title bar of the ACF window 10

11 Section 2) Simple Solution: Second approach is much simpler. Since the input is digital, my only aim was to count the square waves and find the frequency. I have used this alternative because it s a better engineering work from my point of view. As stated, engineers shall produce their products using minimum sources, this approach is exactly meets this statement. A much simpler microcontroller (Pic16F84) can is used. All the calculations took place on the microcontroller. The problem was counting for a period did not seem well, so I have decided to measure the time gap between two heart beats so with a small translation from milliseconds to minutes I have achieved the heart beat rate per minute. Software code for this application is as follows: ;global variables count equ 0x0c ;general purpose counter display3 equ 0x0d ;data for 7-segment display 100's display2 equ 0x0e ;data for 7-segemnt display 10's display1 equ 0x0f ;data for 7-segment display 1's beatlsb equ 0x10 ;heartbeat counter,lsb beatmsb equ 0x11 ;heartbeat counter,msb aargb0 equ 0x12 ;dividend,lsb aargb1 equ 0x13 ;dividend,msb bargb0 equ 0x14 ;divisor,lsb bargb1 equ 0x15 ;divisor,msb remb0 equ 0x16 ;remainder,lsb remb1 equ 0x17 ;remainder,msb timera equ 0x18 ;general purpose timer timerb equ 0x19 ;general purpose timer timerc equ 0x1a ;general purpose timer timerd equ 0x1b ;general purpose timer ;reset vector reset: org 0x00 ;reset vector address goto start ;start program execution start: org 0x06 ;start of program goto Initialisation ; LED DISPLAY DECODER FROM BINARY TO 7-SEGMENT LedDecoder: addwf pcl,f ;W DISPLAY retlw LED_SEGa LED_SEGb LED_SEGc LED_SEGd LED_SEGe LED_SEGf ;0 - "0" retlw LED_SEGb LED_SEGc ;1 - "1" retlw LED_SEGa LED_SEGb LED_SEGd LED_SEGe LED_SEGg ;2 - "2" retlw LED_SEGa LED_SEGb LED_SEGc LED_SEGd LED_SEGg ;3 - "3" 11

12 retlw LED_SEGb LED_SEGc LED_SEGf LED_SEGg ;4 - "4" retlw LED_SEGa LED_SEGc LED_SEGd LED_SEGf LED_SEGg ;5 - "5" retlw LED_SEGa LED_SEGc LED_SEGd LED_SEGe LED_SEGf LED_SEGg ;6 - "6" retlw LED_SEGa LED_SEGb LED_SEGc ;7 - "7" retlw LED_SEGa LED_SEGb LED_SEGc LED_SEGd LED_SEGe LED_SEGf LED_SEGg ;8 - "8" retlw LED_SEGa LED_SEGb LED_SEGc LED_SEGd LED_SEGf LED_SEGg ;9 - "9" retlw LED_BLANK ;10- " " ;Name : Initialisation ;Description : system initialisation and mainloop ;Inputs : none ;Outputs: none Initialisation: ;system initialisation ;port A and port B initialisation bcf status,rp1 ;select bank1 register bsf status,rp0 movlw PORTA_CONFIG1 ;porta configuration movwf trisa movlw PORTB_CONFIG1 ;portb configuration movwf trisb bcf status,rp0 ;select bank0 registers movlw 0x00 ;disable all displays movwf porta ;Name : Mainloop ;Description : measure and display heartbeat ;Inputs : none ;Outputs: none Mainloop: call Heartbeat ;measure heartbeat call Convert ;convert to bpm call Display ;show on display goto Mainloop ;endless loop ;Name : Heartbeat ;Description : measures the time between two heartbeat signals. ; : Triggered on rising-edge of signal, then measures ; : amount of milli-seconds before next rising-edge. ;Inputs : none ;Outputs: beatlsb,beatmsb = length between two heartbeats ; : in milli-seconds Heartbeat: btfss porta,heartbeat_sensor ;wait until signal goes high goto Heartbeat call Delay150 ;wait 150msec - wait until signal has gone low clrf beatmsb ;preset beat registers with 150msec movlw 0x96 movwf beatlsb HL1: call Delay1 ;wait 1msec incf beatlsb,w ;increment beatlsb by 1msec movwf beatlsb 12

13 btfsc status,z ;lsb rollover? incf beatmsb,f ;yes -increment msb btfss porta,heartbeat_sensor ;keep looping till signal goes back high goto HL1 return ;Name : Convert ;Description : convert the measured heartbeat from milliseconds ; : to beats-per-minute. (maybe there's a quicker way!) ;Inputs : beatlsb, beatmsb ;Outputs: display1,display2,display3 = contain hearbeat ; : in bcd format Convert: ;calculate bpm = 60000/beat (lsb and msb) ;Make dividend == 0xEA60 movlw 0xea ;msb movwf aargb1 movlw 0x60 ;lsb movwf aargb0 ;Get measured heartbeat values, as divisor movf beatmsb,w ;msb movwf bargb1 movf beatlsb,w ;lsb movwf bargb0 clrf remb0 ;initialise remainder registers clrf remb1 ; do an unsigned 16-bit by 16-bit division movlw 0x10 movwf count Lu16: rlf aargb1,w rlf remb0,f rlf remb1,f movf bargb0,w subwf remb0,f movf bargb1,w btfss status,c incfsz bargb1,w subwf remb1,f btfsc status,c goto uok16 movf bargb0,w addwf remb0,f movf bargb1,w btfsc status,c incfsz bargb1,w addwf remb1,f bcf status,c uok16: rlf aargb0,f rlf aargb1,f decfsz count,f goto Lu16 ;convert the lsb result of the previous division to unpacked bcd movlw 0x08 ;an 8-bit division movwf count movlw 0x0a ;divide the previous lsb result by 10 movwf bargb0 13

14 clrf remb0 ;initialise remainder Lu8: rlf aargb0,w rlf remb0,f movf bargb0,w subwf remb0,f btfsc status,c goto uok8 addwf remb0,f bcf status,c uok8: rlf aargb0,f decfsz count,f goto Lu8 code!) movlw 0x0a ;3rd bcd digit - 100's is blank(if not changed further down movwf display3 movf remb0,w movwf display1 ;1st bcd digit - 1's movf aargb0,w movwf display2 ;2nd bcd digit - 10's(if not changed further down code!) sublw 0x09 ;is the quotient >=10? bnc $+2 ;yes -quotient is >=10, so divide by 10 again! return ;go back movlw 0x08 ;an 8-bit division movwf count movlw 0x0a ;divide the previous lsb result by 10 movwf bargb0 clrf remb0 ;initialise remainder Lu81: rlf aargb0,w rlf remb0,f movf bargb0,w subwf remb0,f btfsc status,c goto uok81 addwf remb0,f bcf status,c uok81: rlf aargb0,f decfsz count,f goto Lu81 movf remb0,w movwf display2 ;2nd bcd digit - 10's movf aargb0,w movwf display3 ;3rd bcd digit - 100's return ;go back ;Name : Display ;Description : updates display with current measurement ;Inputs : display1-7-segment display 1's ; : display2-7-segment display 10's ; : display3-7-segment display 100's ;Outputs: none Display: movf display1,w ;convert bcd to 7-segment call LedDecoder movwf display1 14

15 movf call movwf movf call movwf display2,w LedDecoder display2 display3,w LedDecoder display3 movlw 0x4 ;update display for a little while movwf timerc DL2: movlw 0xff movwf timerd DL1: clrf porta bsf porta,led_digit1 movf display1,w movwf portb call Delay1 clrf bsf movf movwf call clrf bsf movf movwf call clrf decfsz goto decfsz goto porta porta,led_digit2 display2,w portb Delay1 porta porta,led_digit3 display3,w portb Delay1 porta timerd,f DL1 timerc,f DL2 return ;go back ;Name : Delay ;Description : causes a delay of 1msec ;Inputs : none ;Outputs: none Delay1: movlw 0x02 ;number of wait states movwf timerb L2: movlw 0xa5 ;number of wait states movwf timera L3: decfsz timera,f goto L3 decfsz timerb,f goto L2 return ;go back ;Name : Delay ;Description : causes a delay of 150msec ;Inputs : none ;Outputs: none 15

16 Delay150: movlw 0xc3 ;number of wait states movwf timerb L4: movlw 0xff ;number of wait states movwf timera L5: decfsz timera,f goto L5 decfsz timerb,f goto L4 return ;go back end CHAPTER III - CONCLUSIONS 16

17 First of all I would like to mention that, it is HIGHLY dangerous to play with ECG equipment. You can be subject to unexpected, sudden hazards!.. My final application can be seen in the following picture. Finally I have achieved to apply the auto correlation function on a separate microcontroller. The number (077) seen on the display is the average of the measurements of two different microcontrollers working with the principles, one in section one and the other one as described in section two. FIGURE 3.1 Final application 17

18 During the time, that I was involved in this project, I read a lot, and had the opportunity to learn many things on the subject of heartbeat analysis, which is a starting point for biomedical applications. I have realized that, there exists many other application in biomedical by use of digital signal processing techniques. As a future work I am planning to advance this system by using DSP chips and I would like to have a marketable, robust system. 18

IST TSic Temperature Sensor IC Application Notes ZACwire Digital Output

IST TSic Temperature Sensor IC ZACwire Digital Output CONTENTS 1 TSIC TM ZACWIRE TM COMMUNICATION PROTOCOL...2 1.1 TEMPERATURE TRANSMISSION PACKET FROM A TSIC TM...2 1.2 BIT ENCODING...3 1.3 HOW TO READ