The San Francisco chapter held its April meeting at Dolby Laboratories. Thirty-seven people met to hear Dr. Mei Wu, principal of Mei Wu Acoustics, give a presentation on speech intelligibility.

The goal of speech intelligibility research is to understand how physical environments influence people’s ability to communicate and to predict speech intelligibility of spaces. Wu outlined the evolution of four standards over the past four decades: Speech Interference Level (SIL), Articulation Index (AI), Speech Intelligibility Index (SII), and Speech Transmission Index (STI).

STI measurements make use of a modulation transfer function in which the human voice is characterized by 7 bands of frequencies from 125 Hz to 8 kHz that are modulated in a range of 0.63 Hz to 12.5 Hz. The STI value reveals the depth of modulation. That is, if no noise or reverberation alters the sound, then the speech intelligibility is considered excellent at 1.0. As distortion reduces the depth of modulation, falling STI scores reflect lower speech intelligibility. An STI value of 0 indicates that the communication is totally unintelligible.

Because early speech intelligibility assessment involved time-consuming human listening tests and manual calculations, technology has been developed to automate the measuring process. The first of these was a meter made by Brüel & Kjær in 1985. The current test procedure makes use of a CD containing synthetic voice sounds played through a 3 inch speaker. A meter positioned in various locations in the space is used to measure the STI.

Wu’s case studies involving teleconference rooms and operating/observation rooms have led to following guidelines:

  • A conference room with good speech intelligibility for local conferences may not have good speech intelligibility for teleconferencing.
  • Having a microphone closer to the speaker usually improves speech intelligibility.
  • Adding sound absorptive materials improves speech intelligibility because you reduce the reverberation time.
  • A raised voice helps, but not much in a highly reverberant room.
  • Reducing ambient noise levels improves speech intelligibility, but this may not be in the room designer’s control due to the noise emanating from the people occupying the room.
  • Moving loudspeaker closer to listeners improves speech intelligibility.

Thomas Merklein