IPRO 344: Audio Quality and Energy Efficiency for Mobile Devices and Intercoms

Speech Inteligibility
Earlier in the semester, we decided to go back to using the Speech Transmission Index (STI) instead of the Speech Intelligibility Index (SII). This decision was made because STI is better suited for measuring actual response of an audio system, and SII is more intended for prediction and simulation of sound quality in an environment. The previous software used - LexSTI - was difficult to set up in testing sessions because of difficulties with calibration. One of the goals of this semester was to produce a repeatable, rapid, and simple test procedure that could be applied to any intercom system to develop comparative performance metrics. Jeff Chiles developed a MATLAB script to automate the collection of acoustic data through the sound system. The test procedure was further quickened by the discovery that sound-pressure-level (SPL) calibration, which allows the computer to "know" how loud a sound coming intoit really is, is not necessary to perform an accurate analysis of the STI. As such, the headphone-level sound output was taken directly from the kiosk into a high-quality sound card and recorded using the MATLAB script.


Methdology For Measuring STI
The speech transmission index (STI) has shown to be a valuable tool for the objective assessment the speech intelligibility. From the first presentation in 1971, the method has been refined and diversified for various applications. Recently the International Electrotechnical Commission, IEC, has launched the third revision of the International Standard specifying the method for calculating the index.
STI tests were conducted for a wide range of system configurations, including microphone types, microphone depths, and acoustic foam presence. In addition, the STI was analyzed while a strong source of ambient noise was present. The last series of STI tests were conducted with the intent to develop a visual map of the optimal speaking position for customers communicating in a drive-thru intercom system. The kiosk was moved in incremental distances and rotated in various angles to simulate a customer talking in from different positions, and the STI was recorded for each case.