Figure1.JPG)
A LANGUAGE ACTIVITY MONITOR FOR DIGITIZED SPEECH AAC SYSTEMS TO SUPPORT EVIDENCE-BASED CLINICAL PRACTICE AND OUTCOMES MEASUREMENT
Marvin Indermuhle*, Katya Hill**, Barry Romich*
*Prentke Romich Company
Wooster, Ohio 44691
**Edinboro University of Pennsylvania
Edinboro, Pennsylvania 16444
ABSTRACT
AAC evidence-based practice requires the use of instrumentation to collect data
and measure outcomes (1). The work on automated data logging and the language
activity monitor has been based on the availability of a textual representation
of the language activity, such as is common with text-to-speech based systems.
However, many AAC devices utilize digitized speech and thus have no textual
form of the language activity. This project explored the feasibility of the
notion of LAM for use with digitized speech AAC devices. The Digitized LAM (DLAM)
records the speech output of these devices. The data, which includes time-stamps
and encoded audio, can later be converted to LAM format on a computer and then
analyzed like LAM data.
BACKGROUND
Augmentative and alternative communication (AAC) evidence-based practice and
outcomes measurement emphasize the need for systematic clinical data collection
by teams to support clinical decisions (2). Systematic data collection should
include objective measures of system performance by consumers. Some of the most
valuable objective language measures document performance gathered from language
sampling. The development of the language activity monitor (LAM) has facilitated
language sampling (3). Objective non-language measures involve identifying variables
related to the human-device interface such as access and rate. Objective language
and non-language measures have been reported using automated data logging from
LAM (4). However, current tools and methods involving automated data logging
have been developed only for AAC synthesized speech systems. Consequently, AAC
performance data collection for individuals who rely on digitized systems requires
traditional time consuming methods of observation and recording.
STATEMENT OF THE PROBLEM
Traditional methods of monitoring AAC digitized speech system performance are
based on clinical observation and video or audio recording with subsequent observation,
timing and/or transcription. With the clinical implementation of LAM for AAC
synthesized speech systems, the need to develop tools to support automated data
logging for digitized speech systems became obvious.
RATIONALE
The DLAM development was aimed at providing tools for data collection and analysis
from the speech output of digitized AAC devices. These tools were designed to
provide an easy and accurate method to collect field data on numerous AAC products
currently available from various AAC device manufacturers.
DESIGN
The design was a team effort, including a focus group representing professionals
and consumers with clinical speech-language pathology (SLP), special education,
research, and technical expertise. Design features of the DLAM that were identified
initially included:
· Device compatibility with most digitized AAC products
· Compact flash card to facilitate data storage and memory transfer to a PC
· Internal speaker and external speaker jack for audio interface
· Internal microphone to record an audio cue identifying user or other clinical data
· Real time clock for data time stamp with 1 second resolution
· User interface for record/playback, card storage capacity and other system information
· Compact packaging with rechargeable batteries to accommodate daily use
DLAM functions including the user interface were initially simulated on the PC to refine the physical layout, record/playback functions and other features. Recording routines, system information enunciation, file functions and driver interfaces were developed concurrently with the hardware development.
The DLAM audio input circuit was designed to capture, as accurately as possible, the digitized messages that are reproduced by the various communication devices. This necessitated a connection to the external speaker jack, not a microphone, to eliminate all ambient noise.
DEVELOPMENT
The core components of the DLAM include the WinCE operating system and Intel’s
StrongARM 32-bit RISC (reduced instruction set computer) processor, SA1110.
Using WinCE with the ARM technology provided a standard operating system with
high-performance processing and low-power consumption. This implementation supports
the ATA compact flash card, 16MB of flash memory, 32MB DRAM and FPGA (field
programmable gate array) interfaces.
To minimize interface circuitry for the memory card, real time clock, tone circuitry and other peripherals, a FPGA was implemented. Rechargeable Li-ion batteries were also used in the device because of their high energy density and discharge characteristics. Battery voltage sensing circuitry was implemented to estimate the remaining battery capacity.
The beta DLAMs were packaged in a standard case (FIGURE 1) with the dimensions of 2¾” x 4¾” x 1½”. All user interface functions are located on the top of the device. The compact flash card, switch interface, external speaker connector and serial connector are on the periphery.
DLAMterm is a Windows based software tool developed for the PC to retrieve
the DLAM data and transcribe the recorded language samples. Using this tool,
it is possible to upload DLAM data to a PC, copy saved data from a DLAM memory
card onto a PC, and transcribe the recorded data into a text format compatible
with language analysis software.
This transcription process in the DLAMterm allows the user to step through the
recorded data, one utterance at a time, listen to the utterances, and enter
text representing each utterance. A progress indicator, auto-text completion
and other user-friendly features implemented in DLAMterm simplify the transcription
process.
EVALUATION
The initial phase of this project involved testing the feasibility of the DLAM
under laboratory conditions as well as conducting a reliability study of the
transcription process. In addition, the study involved beta testing under normal
conditions, and consisted of ten field sites located in the following states:
Florida, Louisiana, Mississippi, New Jersey, New York, Ohio, Pennsylvania, and
Virginia. Field testers involved clinicians working in public schools, rehabilitation
centers, developmental centers, and universities.
DISCUSSION
The laboratory and field testing results demonstrated the feasibility of the
technology of a language activity monitor for digitized speech AAC systems in
that encoded audio signals with time stamps can be converted into text. Experience
from the first phase of this project was gained to improve the future direction
of the design and provided useful information in planning for the next phase.
Valuable feedback regarding the design features of the DLAM was received from
the field testers. The use of the flash card posed challenges to the clinicians.
The feedback has resulted in significant simplification of the system controls
and redesign of the DLAM device. In addition, feedback encouraged development
of the DLAM as an internal function, since field testers had experience with
the LAM as a built-in function in synthesized speech AAC devices. The field
testers overwhelming value the concept of language activity monitoring and are
eager to participate in future work. These tools allow for the collection of
data in accordance with the principles of evidence-based practice.
REFERENCES
1. American Speech Language Hearing Association (ASHA) (2001). Scope of Practice.
Rockville, Maryland.
2. Jutai, J., Ladak, N., Schuller, R., Naumann, D., Wright, V., (1996). Outcomes measurement of assistive technologies: An institutional case study. Assistive Technology. 8: 110-120.
3. Romich, B.A. and Hill, K.J. (1999). A language activity monitor for AAC and writing systems: Clinical intervention, outcomes measurement, and research. Proceedings of the RESNA ’99 Annual Conference. Long Beach, CA. pp 19-21.
4. Hill, K.J. and Romich, B.A. (2001). A summary measure clinical report for characterizing AAC performance. Proceedings of the RESNA ’01 Annual Conference, Reno, NV. pp 55-57.
ACKNOWLEDGEMENTS
The National Institute for Deafness and Other Communication Disorders of NIH
has provided funding to Prentke Romich Company to support the work on DLAM.
Marvin Indermuhle,
Prentke Romich Company
1022 Heyl Road
Wooster, OH 44691
Tel: 330-262-1984 ext. 211
Fax: 330-263-4829
Email: mwi@prentrom.com