Presentation 1: Noise canceling in 1-D data

1
Presentation 1 Noise canceling in 1-D data

• Seri Rahayu Abd Rauf
• Fatima Boujarwah
• Juan Chen
• Liyana Mohd Sharipp
• Arti Thumar

M2
2
Overview

• Part of a research done in CMU
• Micron Intelligent Microsurgical Instruments
  project led by Prof Pradeep Khosla
• This design uses adaptive weighted-frequency
  algorithm for noise canceling purposes.

3
Project Background Noise Canceling Chip

• An essential block in a research project done
  here at CMU Micron Intelligent Microsurgical
  Instrument
• Main objective of the research
• To provide a technique to compensate the amount
  of errors due to physiological hand tremor,
  jerk, low-frequency wander and also pathological
  movement disorder.
• To enhance the accuracy of human-machine
  interfaces.
• To improve the living of patients with movement
  disorders.
• To increase performance in microsurgery
  applications.
• Source http//www-2.cs.cmu.edu/camr/research.ht
  ml

m2
4
Applications
Microsurgery Instrument
Rehabilitation
m2
Image sources http//irb.cs.tu-berlin.de/zuse/hi
story/ibm158c.jpg http//www.richard-wol
f.com/english/e_image/ww_hp4.jpg
5
Applications
Vehicle / Aircraft Maneuvering
Hearing Aid
m2
Image sources http//www.ahaanet.com/hearingaidst
yles.asp http//hometown.aol.com/jaywhle
/extreme/control.gif
6
How does the chip work?

• It uses adaptive weighted frequency algorithm in
  1-D data.
• Input Human motion signal with noise
  introduced by pathological and
  physiological tremor.
• Output Noise suppressed signal for
  higher precision.

m2
7
Block Diagram
Source Modeling and Canceling Tremor in
Human-Machine Interfaces
8
Result
Source Modeling and Canceling Tremor in
Human-Machine Interfaces
9
Transistor Estimate
Part Transistors
8-bit Adders 12x8x24 2304
8-bit Multipliers 10x1200 12000
ROM 800
Registers 10x8x14 1160
Misc 3000
Total 19210
10
Status

• Design Proposal v
• Architecture (in progress)
• To be done
• Floor Plan
• Gate Level Design
• Component Layout
• Chip Layout
• Spice Simulation of Entire Chip

11
Problems ?

• Several inputs are too small for bit shifting
• (10-7)
• Value of M could be made greater for higher
  precision (More complex)
• Transistor count seems too simplereally, it is
  NOT

12
Goal

• Higher throughput

13
Alternatives

• Serpent encryption
• Branch predictor

14
Summary

• Why do we choose this?

15
Questions?