Psychoacoustics and Intelligibility

A brief review of psychoacoustics and intelligibility will make it easier to understand the unique and effective approach to ambient noise compensation in Audiostat™ 
Psychoacoustics is the study of the physiology and psychology of the human ear-brain system and its interaction with the acoustics of the listening environment.

Intelligibility is a quantitative measure of the human ability to comprehend speech signals in a given environment. It is expressed as the percentage of the number of words understood correctly to the total number of words.

SmartSpeaker Intelligibilty Test Results
Quantitative intelligibility testing of SmartSpeaker was carried out independently by Mr. Claude Fortier at State of the Art Electroniks, Ottawa. The tests included measurements of Speech Transmission Index (STI) and Percent Articulation Loss of Constants (%ALcons) with and without SmartSpeaker in noisy and quiet environments. A stereophonic recording of real ambient noise was used to create the noise for the field test. The Larson-Davis MLSSA system was used to measure intelligibility.

Signal Source: One 4-inch, 2 Watt loudspeaker producing 65dBA signal SPL at the MLSSA system's test microphone with SmartSpeaker turned off. The test microphone was placed 0.5m from loudspeaker (on axis). Test Signal: Weighted (spectrally shaped) pseudorandom noise generated by the MLSSA system. Ambient Noise: A 15 second segment of a stereophonic recording of subway noise played-back in far-field with stereo reproduction equipment. The noise was composed primarily of human speech (multi-talker babble) with some low-level rumble and hiss from the rolling stock. The background ambient noise level of the test environment was 42dBA, primarily wide-band air-conditioning noise. The SmartSpeaker: Ambient noise pickup microphone was placed 0.2m from the loudspeaker and 0.54m from the test microphone (180o off axis).

General: Power amplifier gain was adjusted for high signal-noise ratio and intelligibility at the test microphone in the absence of ambient noise.

Test Ambient SmartSpeaker Noise % Alcons STI (standard) STI (modified) % Word Intelligibility Rating
1 OFF 42dBA 3.1 0.741 0.929 95.6 Good
1 ON 42dBA 2.5 0.777 0.949 97.3 Excellent
2 OFF 60-64dBA 8.9 0.546 0.662 85.8 Fair
1 ON 60-64dBA 2.1 0.813 0.937 99.5 Excellent
3 OFF 68-72dBA 27.5 0.337 0.393 44.6 Poor
1 ON 68-72dBA 4.2 0.684 0.774 93.5 Good

Where the comprehension of delivered speech is deemed important, intelligibility - not audibility - must define the efficacy of a PA system. Overall, less than 85% intelligibility is considered to be inadequate for most PA applications.

In public environments, the most important cause of intelligibility loss is the interference caused by ambient noise. The impact of ambient noise on speech intelligibility depends in a complex way on several statistical properties of the noise, such as average amplitude, amplitude fluctuation and frequency spectrum etc.

Unlike audibility, intelligibility degradation occurs rapidly when the average noise intensity begins to approach the average PA signal intensity.

The graph shows a typical "inverted S" intelligibility graph. The vertical axis represents intelligibility from 0 to 100%. The horizontal axis represents the ratio of the environmental noise level to the PA signal level on an open-ended logarithmic scale. 

The centre line on this scale indicates when the PA signal level equals the environmental noise level.

On the left of this line, the PA signal level exceeds the environmental noise level, intelligibility increases rapidly and levels off in Section 1.

Further to left, the intelligibility remains high but now the system begins to sound intrusively loud in Section 2 as the PA signal level continues to rise in relation to the level of ambient noise.

Optimum listening conditions lie in Section 1, where the signal is highly intelligible without being intrusively loud.

On the right of the centre line, Section 3 shows a rapid decline in intelligibility which fades into complete inaudibility in Section 4 as the environmental noise level increases in relation to the PA signal level.

Audiostat optimises the PA signal to keep intelligibility in Section 1 - automatically, seamlessly and in real time.