IPRO 343 Places Third in Process and Improvement Track on IPRO Day
by Halcyon Lawrence
"Did they just call my name?"
We all know that announcements in train stations and airports are difficult to understand. This is because public announcements lack standards for talker characteristics. Talker characteristics may include a combination of variables like rate, pitch and duration, language, accents, and gender.
During the Spring 2009 semester, IPRO 343 extended their research begun in the Fall 2008 semester to recommend standards for announcements in noisy and distracting environments. This award-winning IPRO, guided by assistant professor Matthew Bauer (Humanities), investigated three factors that affect speech intelligibility in public announcements: words per minute, pitch, and duration. The team included co-team leaders Shavanna Pinder (ARCH) and Kevin Arnold (SSCI); Jessie Bauer (EECE), Brian Bjerke (EECE), Hyemin Choi (ARCH), Karen Hong (ARCH), Scott Justus (BCPS), Justo Moraga (EECE), Crystal Reynolds (PSYC), and Nor Tanapura (MMAE); and teaching assistant Halcyon Lawrence (Ph.D. student, TECH).
IPRO 343 Team Members [from left]: Prof. Matthew Bauer, Jessie Bauer, Karen Hong, Hyemin Choi, Shavanna Pinder, Kevin Arnold, Crystal Reynolds, Scott Justus, Nor Tanapura, Brian Bjerke, Justo Moraga, Halcyon Lawrence.
Standards and codes exist in almost every built or designed environment--for example, building codes addressing fire safety concerns and accessibility standards for the physically disabled. But there are no standards for talker characteristics on public announcement systems. The lack of standards for talker characteristics introduces the possibility that these important and sometimes life-saving announcements can be misunderstood. The goal of IPRO 343 was to address the lack of standards for talker characteristics in public announcement systems.
Students began the semester by establishing a baseline of public announcements in the Chicagoland area. They made 22 recordings of public announcements in Metra Stations, O'Hare Airport, CTA platforms, and "L" Trains. Then, they analyzed these recordings for talker characteristics including words per minute and duration. Drawing on current research on speech intelligibility, IPRO 343 hypothesized that more intelligible public announcements would be shorter, spoken at a slower rate, and would not be affected by pitch.
Team members [from left] Scott Justus and Brian Bjerke recording public announcements on a CTA platform.
To test their hypotheses, an experiment was designed to measure a participant's accuracy in identifying 40 recorded phrases. Each phrase included a combination of colors, shapes and directions. These phrases were presented to participants at long and short durations, high and low pitches, and slow and fast speeds. Participants were required to listen to the recordings through headphones and indicate what they heard using a software program called Starquiz. Participants' keyboards were customized with stickers representing the colors, shapes, and directions. Customizing the keyboard allowed participant responses to be easily recorded in the form of keystrokes later exported to Excel for analysis.
IPRO 343 experiment in progress.
The team analyzed the data and made the following recommendations about public announcements: longer messages should be broken into shorter messages, or longer messages should be spoken at slower speaking rates; shorter messages should use slow or neutral speaking rates. As expected, pitch had no significant impact on speech intelligibility.
IPRO 343 used a number of innovative methods to recruit 77 participants for their experiment, including establishing an online presence on Facebook (a social-networking Website).
Due to their efforts, the team placed third in the Process and Improvement Track at IPRO Day on May 1, 2009.
IPRO 343 team members responding to questions from a judge on IPRO Day 2009.
posted: