Voice Streaming on a Sensor Network VOSN

1
Voice Streaming on a Sensor Network (VOSN)

• Wei Zhang
• Navraj Chohan

2
Overview

• Hardware Platform
• Software Platform
• Multi-Hop application
• End-to-End application
• Tradeoff
• Future Work
• Questions

3
Hardware Platform

• Motes
• MicaZ
• 2.4GHz, IEEE 802.15.4 compliant
• 250 kbps, High Data Rate Radio
• Wireless Communications with Every Node as Router
  Capability
• ZigBee Compliant
• Laptops
• Communication via serial port
• 57600 baud
• Using MIB510 board
• Connects to mote via UART

4
Software Platform

• TinyOS running on motes
• Routing software
• Endpoints running voice streaming app

5
Firmware Architecture
ReceiveMsg
UARTFramedPacket
6
Multi-hop Routing

• Derived from standard TinyOS route library
• Support 2 way communication
• Remote motes lt--gt base (node 0)
• Each mote support packets buffering and forwarding

7
Multi-hop Path Selection

• Link estimation
• Upstream
• Neighborhood management
• Shortest Path routing protocol ( sp(t) )
• Downstream
• Routing table management
• Routing table lookup

8
Base
9
End-to-End Application

• Serial port communication
• Message input and output queuing
• Packet statistics
• CRC checking
• Text messaging
• Encoding
• Decoding

10
Speex Decoder/Encoder

• Open source
• Written in C
• Used exclusively for voice communication
• Narrow-band (300-3400Hz)
• Wide range of bit-rates (2-44kbps)
• Pick one that suits our needs (8kbps)
• Small voice frames
• 20 ms frames
• 20 bytes per frame (for the bit-rate chosen)
• Squeeze in up to two frames per packet

11
Quality of Service

• Meet minimum Quality of Service (QoS)
• Higher bandwidth usage gives higher QoS
• Send duplicate packets to overcome packet loss
• Getting very good QoS?
• Try to save power at the expense of QoS

12
Power Tradeoffs

• Adaptive power consumption
• When reception is under QoS-low threshold
• Send duplicate packets (raises sent bit-rate)
• When reception is over QoS-high threshold
• No need to send duplicate packets
• Take it a step further by having more frames per
  packet
• More frames per packet
• Cuts down on overhead
• Increases packet size
• Increase in bad CRCs and dropped packets

13
Future Work

• Still much more work to be done
• Use variable bit-rate encoding/decoding according
  to current transmission conditions
• Blur the lines between the end-points and the
  network for optimization
• Power efficient, optimize the overall radio
  transmission under the QoS requirement

14
Questions