Skip to main content

Speech characteristics in coder design

Short answer

Speech signals have limited 300-3400 Hz bandwidth, non-uniform pdf, high sample correlation, non-flat spectrum with formants, pitch periodicity, short-term stationarity over 20 ms frames, and silence periods. Coders exploit these by using 8 kHz sampling, non-uniform quantization, predictive coding (DPCM/ADPCM), LPC modeling, sub-band coding, voice activity detection, and perceptual weighting.

What are different characteristics of speech signals? How are they used in designing of coders?

Characteristics of speech signals

  1. Limited bandwidth – intelligible speech lies mostly in 300–3400 Hz.
  2. Non-uniform probability density – low amplitudes occur far more often than high ones (Laplacian/gamma-like pdf).
  3. Non-zero autocorrelation – adjacent samples (125 µs apart at 8 kHz) are highly correlated (correlation coefficient often > 0.85).
  4. Non-flat spectrum – energy is concentrated at low frequencies, with resonant peaks called formants.
  5. Quasi-periodicity (pitch) – voiced sounds repeat with a pitch period of about 3–12 ms (80–350 Hz); unvoiced sounds are noise-like.
  6. Short-term stationarity – statistics stay nearly constant over 10–30 ms segments.
  7. Silence periods – speakers are silent about 40–60% of the time.
  8. Perceptual masking and phase insensitivity – the ear is relatively insensitive to phase and to noise near strong spectral components.

How these are used in coder design

CharacteristicUse in coder design
Bandwidth 300–3400 HzSample at 8 kHz; filter out other bands
Non-uniform pdfNon-uniform quantization (μ-law/A-law) gives more levels to small amplitudes
Sample correlationPredictive coding (DPCM, ADPCM) sends only the prediction error, fewer bits
Non-flat spectrumSub-band and adaptive transform coding give more bits to strong bands
FormantsLPC models the vocal tract as an all-pole filter; only ~10 coefficients sent
Pitch periodicityLong-term (pitch) predictor; voiced excitation = pulse train
Short-term stationarityProcess in 20 ms frames; update parameters once per frame
SilenceVoice activity detection and discontinuous transmission save power and capacity
Perceptual maskingPerceptual weighting filter in CELP shapes noise under the speech spectrum

Example: the GSM full-rate RPE-LTP coder uses 20 ms frames, an 8th-order short-term LPC filter (formants) and a long-term predictor (pitch) to reach 13 kbps instead of 64 kbps PCM.

✓ Answer checked by Claude Opus 5.5 (high effort)

Related questions

All Wireless Communication chapter 6 questions →