Context-aware%20Sensing%20of%20Physiological%20Signals

1
Context-aware Sensing of Physiological Signals

• Winston H. Wu, Maxim A. Batalin, Lawrence K. Au,
  Alex A. T. Bui, and William J. Kaiser

2
Introduction

• Purpose-Low power consuming physiological sensors
  implementation
• Energy use decreased by enabling disabling the
  sensors to real time measurement demand
• Use low cost sensors to schedule high cost
  sensors like ECG sensors

3
Hardware Used

• Commercially Available PDA with Wifi capabilities
• Bluetooth modules
• 3 Sensors
• ECG sensor
• Pulse Oximeter
• 3 Axis Accelerometer-2 sets

4
Software

• Inference Engine
• GUI
• Local Data Logger
• Device Server
• Device Driver

5
Software Arcitecture
6
Motion detection

• Pulse oximeter used to detect start of the
  exercise
• 2 Accelerometers used to detect end of the
  exercise
• 1 on right ankle and 1 on left hip
• Inference engine on the wearable system computes
  when to activate ECG sensor
• Data collected is streamed to a central server
  via Wifi Network

7
Communication Via Bluetooth

• Each data point accompanied by tracking sequence
  number to check for errors
• PDA is the master node over bluetooth network

8
Context Aware Sensing

• Feature Extraction
• Pulse rate and SpO2 value-rate of decline of
  oxygen saturation
• Accelerometer
• Since cyclical movements are involved
• Features from spectral domain are used
• In general case features from time domain may be
  used
• 512 data points window-100 points entered every
  second
• 2 spectral feature values extracted from each
  axis -f peak and f energy

9
Context Aware Sensing Algorithm of an ECG signal
10
Naïve Bayes Classifier

• P(C/F)
• Where C is the patient states of interest
• F is the feature vector
• Pulse classification as Low , Medium, High
• When high Accelerometer activated
• Accelerometer classifies as Rest , Walk, Jog, Run
• If Jog or Run ECG sensor not activated
• Else it is activated

11
Results
12
THANK YOU