Speech Reinforcement: Human Speaker

By: Owen T. Heisler
Updated: 2019-10-06
Published: 2016-08-05

This article is part of a series; see Speech Reinforcement.

The human behind the microphone is obviously the most important part of the signal path. A well designed sound system can easily accommodate most voices.

Suggestions for the person speaking

It is seldom practical to try altering the behavior of the person speaking, and generally impossible to affect a change in the actual sound produced. However, there are a few suggestions that could be shared with any speaker who is interested.

  1. “Speak up!” If the speaker is aware that additional effort is needed, it can actually cause involuntary changes in speech for improved clarity (Lombard effect), in addition to a simple signal-to-noise ratio increase at the microphone.
  2. It can be counter-productive for the lips to be closer to the microphone than about 3 inches. This is caused by the proximity effect, which increases low frequency gain and thus reduces the signal-to-noise ratio. The speaker should move closer to the microphone only when the system has feedback or gain problems (it should not be necessary with correct system design).

Frequency ranges

Having covered some suggestions for the speaker, we now are concerned with the technical characteristics of voice so we can understand what our “signal” actually is.

  • Typical adult male speech fundamental frequency: 85 Hz to 180 Hz.
  • Typical adult female speech fundamental frequency: 165 Hz to 255 Hz.
  • Child’s speech fundamental frequency: 250 Hz to 300 Hz or higher.
  • Maximum vocal range: 65 Hz (male) to 1.28 kHz (female).
  • Sibilance, or essing, is the high frequency sound associated with the letter s. Sibilants range from 2 kHz to 10 kHz.
  • Speech critical range: approximately 170 Hz to 8.3 kHz.
  • Range for high-fidelity speech reproduction: 80 Hz to 12 kHz.
  • In telephony, the usable voice frequency range is approximately 300 Hz to 3.4 kHz. This is not adequate for high-fidelity speech reproduction; the fundamental frequency of most speech and much of the sibilance of speech falls outside of this range.
  • As comparison, a standard piano keyboard covers tones from 27.5 Hz to 4.186 kHz.

Sources:

Speech critical bands

The following is a selection of narrow frequency bands which may be treated as the elementary signals in speech. This is from Gdansk University of Technology Multitask Noisy Speech Enhancement System.

Speech critical bands

Center

Width

Range (Hz)

200

60

170-230

300

60

270-330

500

60

470-530

800

70

765-835

1000

80

960-1040

1500

100

1450-1550

2000

130

1935-2065

3000

200

2900-3100

5000

300

4850-5150

8000

600

7700-8300

Consequently, on a 1/3 octave equalizer, the following bands are especially important:

  • 160 Hz
  • 200 Hz
  • 250 Hz
  • 315 Hz
  • 500 Hz
  • 800 Hz
  • 1 kHz
  • 1.6 kHz
  • 2 kHz
  • 3.15 kHz
  • 5 kHz
  • 8 kHz

Interestingly, all other bands can be reduced quite a lot without noticably impacting audio quality. However, I do not recommend doing so for normal operation, as everything from 80 Hz to 10 kHz is important for high-fidelity speech reproduction.

I share freely-licensed articles and open-source software. Please consider supporting my work if you find it useful. Thank you!