1. Overview BWSSN Recap Skin Sensor Signal Types DSC Brief Review
DWT Brief Review
ECG Signal
DWT ECG
DWT ECG Adapted for Wireless Transmission
DSC DWT ECG Wireless Transmission

2. Typical Skin Sensor Network
Each of these sensors will be modified and adapted for wireless communications
The long term plan is that once operation and performance is optimized, devices will be miniaturized
It is intended that these devices be attached to the skin and worn continuously say for periods of a few days maybe longer as power performance improves
Trend of increasing memory and processing power and reduced power consumption in the remote devices

3. BWSSN Basic Sensor Processor Module
Modules will incorporate sensing, signal processing, communications and low power electronics.
Test bed system will initially comprise the crossbow motes interfacing with USF developed and off the shelf skin sensors for
Temperature
Heart Rate
Blood Oxygen
Blood Pressure
Sweat component detectors

4. Skin Sensor Signals
Modules will incorporate sensing, signal processing, RF communications and low power electronics.
Skin sensors for
Temperature: Thermistor
Heart Rate: ECG
Blood Oxygen: Pulse Oximeter
Blood Pressure
Sweat component detectors:

5. Distributed Source Coding (DSC) Typical Area of Application
Wire- less
Comm
network
Central
Decoder X Y2 Y1 Y3 Y4
Y1, Y2,.. Represent multiple correlated sources
How to utilize this correlation between sources to reduce bit transmissions without losing information at the decoder
Typically, designate one reference signal from one source and use the others to send error check bits or portions of their original signal to reinforce or confirm the reference signals accuracy.
Yi = hi*X + Zi
Zi: Noise
hi: LTI Filter

6. DSC SW Example Implementation00
X=000
Source X
000
Encoder
Decoder
000
001
010
011
100
101
110
111
{000, 111} = 00
{001, 110} = 01
{010, 101} = 10
{011, 100} = 11
X={000, 111}
Y=000
dH(X,Y) ≤1
X-Y ≤1
X=000
Transmitted bits have been reduced from 3 per symbol to 2 per symbol without loss of information at the receiver

7. Wireless Communication General Data Packet Structure
Preamble sequence
Start of Packet Delimiter
PRE
SPD
LEN PC
ADDRESSING
DSN
Link Layer PDU
CRC
CRC-16
Data sequence number
Addresses according to specified mode
Flags specify addressing mode
Length for decoding simplicity

8. TOSH Data Packet Used in Xbow Motes
A typical TOSH data packet comprises the following:
7E 42 FF FF D B9 07 B0 07 BE 07 B5 07 7F 00 FF 01 FF E
In this frame:
7E 42 ==> SYNC and TYPE bytes is the Serial framing protocol
The rest is the TOS_msg -- actual packet (detailed in file tos/types/AM.h)
Payload in blue: 29 actual data bytes

9. TOSH Data Packet Breakdown is as follows: uint16_t addr; // FF FF
uint8_t type; // 06 //
uint8_t group; // 11
uint8_t length; // 1D
int8_t data[TOSH_DATA_LENGTH]; // B9 07 B0 07 BE 07 // B5 07 7F 00 FF 01 FF //
uint16_t crc; // 55 86

10. Energy Consumption Reduction Via DSC
TOSH_DATA_LENGTH is usually set equal to 29. Varying this number varies the length of the data packet.
Each packet carries node id, voltage, temperature, light and other mote information.
Each sensor’s data is encoded into two eight bit words.
One objective via DSC
reduce the number of data bits per sensor that is wirelessly transmitted from 16 to 8 with the receiver still yielding the sensor reading without loss of information.
E.G. Tx of temperature. All sensors are in the region of 34 deg C. say S1=34.5, S2=34.7, S3= Tx only the information after the decimal with one reference say S1 of 34.5 deg C.

11. Wavelet Transforms Introduction
A small wave
Wavelet Transforms
Convert a signal into a series of wavelets
Provide a way for analyzing waveforms, bounded in both frequency and duration
Allow signals to be stored more efficiently than by Fourier transform
Be able to better approximate real-world signals
Well-suited for approximating data with sharp discontinuities
“The Forest & the Trees”
See gross features with a large "window“
See small features with a small "window”
A waveform that is bounded in both frequency and duration. Wavelet transforms provide an alternative to more traditional Fourier transforms used for analysing waveforms, e.g. sound.
In theory, signals processed by the wavelet transform can be stored more efficiently than ones processed by Fourier transform. Wavelets can also be constructed with rough edges, to better approximate real-world signals.
If we look at a signal with a large "window," we would notice gross features. Similarly, if we look at a signal with a small "window," we would notice small features. The result in wavelet analysis is to see both the forest and the trees, so to speak
This makes wavelets interesting and useful.
The wavelet analysis procedure is to adopt a wavelet prototype function, called an analyzing wavelet or mother wavelet. Temporal analysis is performed with a contracted, high-frequency version of the prototype wavelet, while frequency analysis is performed with a dilated, low-frequency version of the same wavelet.
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

12. TIME-DOMAIN SIGNAL The Independent Variable is Time
The Dependent Variable is the Amplitude
Most of the Information is Hidden in the Frequency Content
10 Hz
2 Hz
20 Hz
2 Hz +
10 Hz +
20Hz
Time
Magnitude
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

13. STATIONARITY OF SIGNAL
Time
Magnitude
Frequency (Hz)
2 Hz + 10 Hz + 20Hz
Stationary
Occur at all times
Do not appear at all times
Time
Magnitude
Frequency (Hz)
Non-Stationary
: 2 Hz +
: 10 Hz +
: 20Hz
ECG signal is non stationary
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

14. Multiresolution Analysis
Wavelet Transform
An alternative approach to the short time Fourier transform to overcome the resolution problem
Similar to STFT: signal is multiplied with a function
Multiresolution Analysis
Analyze the signal at different frequencies with different resolutions
Good time resolution and poor frequency resolution at high frequencies
Good frequency resolution and poor time resolution at low frequencies
More suitable for short duration of higher frequency; and longer duration of lower frequency components
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

15. PRINCIPLES OF WAVELET TRANSFORM
Split Up the Signal into a Bunch of Signals
Represents the Same Signal, but all Corresponding to Different Frequency Bands
Only Providing What Frequency Bands Exists at What Time Intervals
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

16. PRINCIPLES OF WAVELET TRANSFORM
Basically WT is a convolution of the wavelet function with the signal
WT analyzes signal info by modifying the wavelets thru translation (location) and dilation (scale)
Wavelet function φa,b with Scale, a and Location, b
T. Froese ECG Signal Classification using DWT:

17. DEFINITION OF CONTINUOUS WAVELET TRANSFORM
Translation
(The location of the window)
Scale
Mother Wavelet
Wavelet
Small wave
Means the window function is of finite length
Mother Wavelet
A prototype for generating the other window functions
All the used windows are its dilated or compressed and shifted versions
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

18. Translated 1 D Wavelet Example
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

19. DWT SCALE
Scale
S>1: dilate the signal
S<1: compress the signal
Low Frequency -> High Scale -> Non-detailed Global View of Signal -> Span Entire Signal
High Frequency -> Low Scale -> Detailed View Last in Short Time
Only Limited Interval of Scales is Necessary
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

20. Time Scaled 1D Wavelet
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

21. MATHEMATICAL EXPLANATION
CWT can also be regarded as the inner product of the signal with a basis function
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

22. COMPUTATION OF CWT
Step 1: The wavelet is placed at the beginning of the signal, and set s=1 (the most compressed wavelet);
Step 2: The wavelet function at scale “1” is multiplied by the signal, and integrated over all times; then multiplied by ;
Step 3: Shift the wavelet to t= , and get the transform value at t= and s=1;
Step 4: Repeat the procedure until the wavelet reaches the end of the signal;
Step 5: Scale s is increased by a sufficiently small value, the above procedure is repeated for all s;
Step 6: Each computation for a given s fills the single row of the time-scale plane;
Step 7: CWT is obtained if all s are calculated.
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

23. RESOLUTION OF TIME & FREQUENCY
Higher Frequencies
Time
Frequency
Better time resolution;
Poor frequency resolution
Better frequency resolution;
Poor time resolution
Each box represents a equal portion
Resolution in STFT is selected once for entire analysis
Lower Frequencies
Regardless of the dimensions of the boxes, the areas of all boxes, both in STFT and WT, are the same and determined by Heisenberg's inequality . As a summary, the area of a box is fixed for each window function (STFT) or mother wavelet (CWT), whereas different windows or mother wavelets can result in different areas. However, all areas are lower bounded by 1/4 \pi . That is, we cannot reduce the areas of the boxes as much as we want due to the Heisenberg's uncertainty principle. On the other hand, for a given mother wavelet the dimensions of the boxes can be changed, while keeping the area the same. This is exactly what wavelet transform does.
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

24. Discrete Wavelet Transform
Discrete form of Wavelet function
m varies dilation,
n varies translation,
a0 is a fixed dilation step,
b0 is the location parameter > 0
Natural way to sample parameters a and b is to use a log discretization of the ‘a’ scale and link this in turn to the size of the steps taken between ‘b’ locations
T. Froese ECG Signal Classification using DWT:

25. Discrete Wavelet Transform
DWT of a continuous signal via a discrete wavelet is given by
Tm,n are known as wavelet or detail coefficients
In the dyadic grid arrangement a0=2, and b0=1
Discrete dyadic wavelets are usually selected to be orthonormal (orthogonal and normalized to have unit energy)
Means that info stored in a wavelet coefficient, Tm,n is not repeated elsewhere and allows for complete regeneration of the original signal without redundancy
T. Froese ECG Signal Classification using DWT:

26. Discrete Wavelet Transform
Sampling of the continuous signal is accomplished by associating the orthonormal dyadic discrete wavelets with scaling functions
Scaling function is convolved with the continuous signal to produce approximation coefficients which are weighted averages of the signal factored by 2m/2
These approximation coefficients Sm,n constitute a discrete approximation of the signal at scale m
T. Froese ECG Signal Classification using DWT:

27. Continuous Wavelet Transform Discrete Wavelet Transform
T. Froese ECG Signal Classification using DWT:

28. 1d DWT Algorithm
Starting from s, 1st step produces 2 sets of coeff: Approx coeff. cA1 and Detail coeff cD1. These vectors are obtained by convolving s with LPF Lo-D for cA1 and with HPF Hi-D for cD1 followed by dyadic decimation
Source: Matlab

29. Matlab DWT Decomposition Algorithm
Source: Matlab

30. DWT Matlab Implementation
Source: Matlab

31. SUBBAND CODING ALGORITHM
Halves the Time Resolution
Only half number of samples resulted
Doubles the Frequency Resolution
The spanned frequency band halved Hz
D2: Hz
D3: Hz
Filter 1
Filter 2
Filter 3
D1: Hz
A3: Hz A1 A2
X[n]512
256
128 64 S D2 A3 D3 D1

32. Matlab Illustration
Source: Matlab

33. Matlab Wavelet Decomposition
Source: Matlab

34. Matlab Wavelet Decomposition
Source: Matlab

35. WAVELET BASES Wavelet Basis Functions: Time domain Frequency domain
Derivative Of a Gaussian
M=2 is the Marr or Mexican hat wavelet
Time domain
Frequency domain

36. Various 1D Wavelets
Source: Fengxiang Qiao, Ph.D. Texas Southern University
Introduction%20to%20Wavelet%20a%20Tutorial%20-%20Qiao.ppt

37. Wavelet Function Criteria
For complex wavelets, FT must both be real and vanish for negative frequencies
Wavelet must have finite energy
Wavelet must have zero mean meaning that wavelet has no zero frequency components

38. DWT of ECG
DWT can be used to represent a discretely sampled ECG signal by a finite amount of time-invariant wavelet coefficients with enough resolution for further signal diagnostics.
DWT can also be used to filter noise and other signal deteriorating artifacts by dropping out selected coefficients in the reconstruction process

39. ECG Signal Origin
Electrical / Biological schematic of heart conduction system
SA node discharges spontaneously on it own even in a test tube

40. ECG Signal Origin
Heart electrical signal propagation schematic – gives rise to the ECG signal detected on via the skin surface
SA node originates signal, AV regenerates pulse. In response to AV regenerated pulse, electrical discharge across the ventricle muscles causing contraction
Cardiac conduction system: The specialized network of cells in the heart that initiates an electrical signal in the heart and carries it throughout the heart, causing it to beat.

41. Standard ECG Signal Detection
Standard reference measurement of three leads fro ECG signal detection.
Get a signal from each of the three leads

42. Typical ECG Signal Ref: http://rnbob.tripod.com/index.htm
Standard ECG signal showing P-Q-R-S-T-U points characterizing ECG signal
Ref:

43. ECG Signal Points of Origin
                                                    [Diagram from Psychophysiology-Human Behaviour and Physiological Response
Points on ECG waveform related to physical heart electrical signal
[Diagram from Psychophysiology-Human Behaviour and Physiological Response]

44. ECG Wave Components
Formal definition of points and segment of the ECG signal
Each component segment termed a wave
R-R interval or P-P interval constitutes a beat
The P wave represents atrial activation; the PR interval is the time from onset of atrial activation to onset of ventricular activation. The QRS complex represents ventricular activation; the QRS duration is the duration of ventricular activation. The ST-T wave represents ventricular repolarization. The QT interval is the duration of ventricular activation and recovery. The U wave probably represents "afterdepolarizations" in the ventricles.

45. Terminology for Contraction Phases
Marquette Electronics Copyright 1996,

46. Note that each PQRSTU wave does not repeat at exactly the time wrt to the previous.
Each PQRSTU wave appear to have similar duration, but may not start at the same time each time
PR interval: .12 to .2 seconds, QRS interval <.12 secs.

47. < 60 bpm

48. Fast heart rate

49. Slow heart rate

50. Abnormal heart rate: heart attack

51. Ventricles are blocked or not contracting in response to AV excitation

52. Marquette Electronics Copyright 1996

53. Heart Rate Variability - HRV
To perform geometric measures on the NN interval histogram, the sample density distribution D is constructed, which assigns the number of equally long NN intervals to each value of their lengths. The most frequent NN interval length X is established, that is, Y=D(X) is the maximum of the sample density distribution D. The HRV triangular index is the value obtained by dividing the area integral of D by the maximum Y. When the distribution D with a discrete scale is constructed on the horizontal axis, the value is obtained according to the formula HRV index=(total number of all NN intervals)/Y. For the computation of the TINN measure, the values N and M are established on the time axis and a multilinear function q constructed such that q(t)=0 for t N and t M and q(X)=Y, and such that the integral 0+ (D(t)-q(t))2 dt is the minimum among all selections of all values N and M. The TINN measure is expressed in milliseconds and given by the formula TINN=M-N. Also see Table 1.
Standards of Measurement, Physiological Interpretation, and Clinical Use
Task Force of the European Society of Cardiology the North American Society of Pacing Electrophysiology. Correspondence to Marek Malik, PhD, MD, Chairman, Writing Committee of the Task Force, Department of Cardiological Sciences, St George's Hospital Medical School, Cranmer Terrace, London SW17 0RE, UK.

54. Normal Values of Standard Measures of HRV
Variable
Units
Normal Values (mean±SD)
Time Domain Analysis of Nominal 24 hours
SDNN ms
141±39
SDANN
127±35
RMSSD
27±12
HRV triangular index
37±15
Spectral Analysis of Stationary Supine 5-min Recording
Total power
ms2
3466 ±1018 LF
1170±416 HF
975±203 nu
54±4
29±3
LF/HF ratio

55. Typical Autonomic Response (ANS) Affecting Heart Rate
[Diagram From The Institute of HeartMath]

56. Typical ECG Signal Noise and Distortions

58. Clean ECG ECG+ noise+60hz DWT Filtered Signal NIECG1 is in slide
NIEGCG shows multiple mother wavelets
DWT Filtered
Signal

59. Sym5 Processed Signal Clean ECG ECG+ noise+60hz DWT Filtered Signal
Sym5 dwt
DWT Filtered
Signal

60. Db5 Processed Signal Clean ECG ECG+ noise+60hz DWT Filtered Signal
Db5 dwt
DWT Filtered
Signal

61. Detected for signal recovery
Sym5 Processed Signal
Clean
ECG
ECG+ noise+60hz
DWT Filtered
Signal
Sym5 Processed Signal
Peaks
Detected for signal recovery

62. Detected for signal recovery
Db5 Processed Signal
Clean
ECG
ECG+ noise+60hz
DWT Filtered
Signal
Db5 processed signal
Peaks
Detected for signal recovery

63. Matlab DWT Threshold Selection Choices
Matlab DWT Denoising function
[XD,CXD,LXD] = wden(X,TPTR,SORH,SCAL,N,'wname')
Wavelet decomposition is performed at level N and 'wname' is a string containing the name of the desired orthogonal wavelet
TPTR: Threshold Selection Rules:
'rigrsure' use the principle of Stein's Unbiased Risk
'heursure' is an heuristic variant of the first option
'sqtwolog' for universal threshold
'minimaxi' for minimax thresholding (see thselect for more information)
SORH ('s' or 'h') is for soft or hard thresholding (see wthresh for more information).

64. Multiplicative Threshold Rescaling
SCAL defines multiplicative threshold rescaling:
'one' for no rescaling
'sln' for rescaling using a single estimation of level noise based on first-level coefficients
'mln' for rescaling done using level-dependent estimation of level noise

65. DWT ARHOS LP Filter
R.H. Istepanian, L.J. Hadjileontiadis, S.M Panas, ECG Data Compression Using Wavelets and Higher Order Statistics Methods, IEEE Transactions on Information Technology in Biomedicine, Vol 5, No.2 June 2001

66. Summary Research Goals
Signal processing algorithms for wireless skin sensors with the following optimization features and objectives
Higher signal to noise output with lower input signal power
Common interference signals identified and cancelled
Common and typical artifacts characterized, reduced and / or cancelled
Reduced instruction set, software memory and hardware requirements for processing of signal
-DSP based techniques
- Filtering, Wavelet, Empirical Mode Decomposition

67. Summary Research Goals
Utilization of multiple wireless sensors to improve desired signal via constructive signal addition from the multiple sensor outputs.
Processor hardware and software designed and optimized for low power consumption
Sensors to include types for measuring pulse, respiration, blood oxygenation, glucose levels, bio-impedance, skin hydration, and body temperature.
Majority vote basis for signal received decisions
Hardware software optimization Like tinyos mote implementation