1. EPS Sensors & Hand Tracking/Gesture Recognition
IT-S0C Lab. Chonnam National Univeristy

2. Contents System Summary EPIC Sensor: Circuits & Principles
Source Code Analysis Using Labview
NUI-Based Data Preprocessing
NUI-Based Gesture Recognition Algorithm

3. Chapter 1
System Summary

4. System Summary Human body outside charge ↑ ⇒ outsde voltage ↑
(Dielectric area)
Sensor probe area
EPIC sensor
Input
: Dielectric permittivity bet.
human and sensor
Electrostatic Field
: permittivity of dielectric material
deposited on EPS sensor
Electric field surrounding EPIC sensors

5. Chapter 2
EPIC Sensor: Circuits & Principle

6. EPIC Sensor: Circuit Analysis (1/2)
Electric Potential Integrated Circuit
EPIC sensor: data detection principle
■ Detect surrounding E-field variation
■ Non-directional & metal-shut off
Contact mode : bio-electric signal (ECG, EMG, EOG, etc.) measurement
Non-contact mode : Measure variation
in surrounding E-field
Single-ended mode :
measure electric potential
Differential mode : measure
difference bet. two sensors
- Main detected signal : power line
noise
■ Variation in static-electric charge
Once objects with different dielectric constants) enter E-field, static-electric charges change
■ Disturbance in E-field
Utilize the principle that disturbance in E-filed occurs due to the movement of human body whose role is like big container with polarity
EPIC sensor input stage
Bootstrapping : Corner frequency control(input resistance)
Guarding : Gain control(input impedence)
Basic block circuit diagram of EPIC sensor
Contact mode & Non-contact mode

7. Input stage of a EPIC sensor
EPIC Sensor: Circuit Analysis (2/2)
Input stage of a EPIC sensor

8. Simulation for verification
Simulation 2.

9. Simulation for verificaton
Non Inverting OP-AMP
Kirichhoff’s current Law on node x
Vx (virtual short voltage) = Vi(input voltage).
- Virtual short:
Input node voltage feedback
AMP
equals to voltage of inverting node voltage
Non-inverting OP-AMP

10. Simulation for Verification
Input bias current
Current for OP-AMP operation
Feed back loop gain becomes to be different from open loop gain, but it generate Offset
In order to solve “offset” problem, resistance, Rin, is connected to outside of OP-AMP
Input stage of EPIC sensor

11. Simulation for Verification
Simulation Result(1/2)
Frequency : 1000Hz
C_ext : 250pF, C_in : 15pF, Rin : 20GΩ, Av = 50(PS25401)
Cext = 250pF, Freq = 1000 Hz
RG1/RG2 = 52, RG1 = 0.01~100GΩ
Cext = 250pF, Freq = 0~100kHz
RG1/RG2 = 52, RG1 = 1GΩ

12. 이론 검증을 위한 시뮬레이션 시뮬레이션 결과(2/2) 주파수 : 1000Hz
C_ext : 250pF, C_in : 15pF, Rin : 20GΩ, Av = 10(PS25405)
Cext = 250pF, Freq = 1000 Hz
RG1/RG2 = 9.6, RG1 = 0.01~100GΩ
Cext = 250pF, Freq = 0~100kHz
RG1/RG2 = 9.6, RG1 = 1GΩ

13. EPIC Detection Principle & Analysis (1/2)
Change in static electric charges
Simulation result
Sensor Electrode
Real output from EPIC
Electrometer
Amplifier
Skin
Air
Condition : PC, CIB grounding & surrounding E-field V/m

14. EPIC Sensor Detection Principle & Analysis(2/2)
Disturbance in E-field
PS25201 EPIC voltage due to movement of a target
PS25201 EPIC voltage standard deviation according to distance of a moving target
Condition : PC, CIB grounding & surrounding E-field V/m

15. Chapter 3
Source Code Analysis using Labview and C++ Compatibility Method

16. Computer based measurement system (DAQ hardware equipped
Source code anlaysis using Labview tool(1/2)
DAQ (Data Acquisition) Process
DAQ hardware : Measure and generate electric signals
In general, 10 AD conversion per period is desirable
For example, set sampling rate to more 100KS/s if 10KHz is measured
AI : Analog signal measurement
AO : Analog signal generation
DI : Digital signal measurement
DO : Digital signal generation
CI : Pulse edge no. measurement
CO : Pulse signal generation
Computer based measurement system
(DAQ hardware equipped

17. Source Code Analysis Using Labview tool(2/2)
DAQ (Data Acquisition) Process
Measure continuous signal measurement using hardware timing
Channel generation
Analog input > voltage
Timing setting
Sample mode > continuous sample
Data reading
analog > Multiplexed channels > multiplexed samples > 1D Wfm(waveform)
Generate the channel
Setting the timing
Start
Read the data
Stop
Clear the data
Error handle

18. C++ 호환 방법 C++ tool 을 이용한 프로그래밍을 위한 준비 사항 드라이버 설정 펌웨어 업데이트
해더 파일과 라이브러리 파일 설정
“NIDAQmxBase.h”, “NIDAQmxBase.lib”
DAQ 과정에 따른 DAQmxErrChk API 함수 사용
DAQmxBaseCreateTask / DAQmxBaseCreateAIVoltageChan /
DAQmxBaseCfgSampClkTiming / DAQmxBaseCfgInputBuffer /
DAQmxBaseStartTask / DAQmxBaseReadAnalogF64 / DAQmxBaseGetExtendedErrorInfo /
DAQmxBaseStopTask

19. Chapter 4
NUI Based Data Preprocessing

20. IIR Low Pass Filter Source Codes
IIR LPF 2st 계수 설정

21. Data Preprocessing Process(1/2)
정전기 잡음
목표물의 움직임으로 발생된 신호
[ extracted signals in case that static electricity occurs]
[ Variation in frequency domain according to movement]
[ signal variation caused by static electricity]

22. Data Preprocessing Process(2/2)
[Data preprocessing result in case of a fixed target]
[ Data preprocessing result in case of a moved target]

23. Chapter 5
NUI Based Gesture Recognition Algorithm

24. [Flow diagram of the proposed DTW ]
NUI Gesture Recognition Based DTW Algorithm
IIR Low-pass filter
10Hz 2st order(AC)
Non-Contact
Electrometer Sensor Data
Analog voltage(AC)
Generation of
Electrostatic charge?
True
True
Calibration enabled ?
Data preprocessing
False
Substitution to previous data buffers
False
Extracting maximum
Value of data buffers
Digital voltage(DC)
Timer1 enable
Extracting
differential signals
Store of voltage
Calibration processing
Kalman filter
False
Timer1 = 3sec?
False
Velocity > Th_vel ?
True
Extracting VNH and Th_vel
True
Timer2 enable
Calibration enable
Store of voltage
False
Timer2 = 1sec?
True
Training data
Data normalization
DTW Runtime recognizer
Classification
Argmin(class)
Event
[Flow diagram of the proposed DTW ]

25. 구현된 Warping 구간의 전역 상수 제한 범위
NUI Gesture Recognition Based DTW Algorithm
DTW(Dynamic Time Warping) algorithm
Algorithm measuring similarity between two data in terms of time change
Comparing simultaneously by searching optimal non-linear mapping function
구현된 Warping 구간의 전역 상수 제한 범위

26. Warping Path Type

27. THANK YOU
FOR YOUR ATTENTION
Any Questions ?