Digital Voice Waveform Selection

1
Digital Voice Waveform Selection

• J. W. Nieto
• Harris Corporation
• RF Communications Division

2
Presentation Overview

• Presentation Overview
• Digital Voice
• Operational Requirements
• Waveform Requirements
• HF Waveform Standards
• DV Performance
• Waveform Performance
• Recommendation

3
Digital Voice

• Starting Point
• 8000 samples per second
• 12 or 16 bits per sample
• Compression (u-law, a-law) yields 64 kbps
• 8 bits per sample
• Continuously Variable Slope Delta (CVSD) yields
  16000 bps
• Lower bit rates require modeling characteristics
  of human voice

4
Digital Voice

• Linear Predictive Coders (termed vocoders)
• Human voice specific (model human voice tract)
• High compression of speech
• Low quality of non-speech signals
• Periodic pulse train or white noise excite an all
  pole synthesis filter
• Sometimes sound mechanical or buzzy
• Prone to annoying thumps and tonal noises
• LPC10e -gt 2400 bps

5
Digital Voice

• Mixed Excitation Linear Prediction (MELPe)
• Funded by US Government
• Based on traditional LPC10e parametric model
• Additional Enhancements
• Mixed excitation
• Aperiodic pulses
• Pulse dispersion
• Adaptive spectral enhancement
• Fourier magnitude scaling
• Noise preprocessor
• 2400 bps and 1200 bps
• Harris proprietary 600 bps

6
Digital Voice

• Desired Digital Voice Characteristics
• Intelligibility
• Voice Quality
• Speaker recognition
• Communicability
• Robust to acoustic background noise environments
• Low data rates
• Robust to bit errors
• Low computational complexity
• Low latency
• MELPe significantly better than LPC10e

7
Vocoder Parameters
8
Operational Requirements

• Secure communications
• Integrated voice and data (with quality of
  service)
• Interoperability
• Long battery life
• Small light-weight radio
• Automated functionality (i.e. Automatic Link
  Establishment, hands free operation)
• Covert and easy to deploy antenna
• Automated selection of DV algorithm and physical
  layer waveform based on channel conditions
• Ad-hoc networking
• Broadcast, conference and private communications
• Good performance in difficult acoustic background
  noise environments

9
Waveform Requirements

• Robust preamble able to provide error-free
  Message Indicator (MI) and other required
  information (data rate, interleaver size, traffic
  type, etc)
• Sound quality mostly immune to fading and
  interference common to HF communications
• Handle wide range of multipath/fading conditions
• Low latency
• Handle narrow band interference and impulsive
  noise
• Small peak-to-average ratio (PAR) (i.e. maximize
  transmit power)
• Reduced adjacent channel interference

10
HF Waveform Standards
11
STANAG 4539
12
DV Random Bit Error Performance
13
Waveform Performance
PDF Mid-Latitude Disturbed Channel Condition
14
Waveform Performance
PDF Mid-Latitude Quiet Channel Condition
15
Waveform Performance
CDF Mid-Latitude Disturbed Channel Condition
16
Waveform Performance
CDF Mid-Latitude Quiet Channel Condition
17
Recommendation

• Standardized HF waveform to use for Digital Voice
• STANAG 4539
• 2400S, 1200S, 600S
• Provides autobaud and automode (data or DV)
• Integrated voice and data system
• Issues with developing new waveforms designed
  specifically for vocoders
• Costly development and standardization process
• Which bits should be coded ?
• Which FEC most effective on HF multipath fading
  channels ?
• How many different HF channel conditions should
  be tested ?
• How many hours of speech should be tested to
  select best waveform ?
• Would it be significantly better than STANAG 4539
  ?