Flexible Integration of Voice Recognition Components for an Automotive Android Platform: A Design Science Research
Typ
Examensarbete för masterexamen
Program
Publicerad
2020
Författare
Johansson, David
Eliasson, Marcus
Modellbyggare
Tidskriftstitel
ISSN
Volymtitel
Utgivare
Sammanfattning
There are several voice recognition interfaces available on the market capable of
being integrated into a variety of platforms. Mobile phones, speakers, and even
refrigerators can be equipped with a voice assistant. Having a fully integrated voice
assistant in a car has been shown to increase the performance of the driver and reduce
distractions. The subcontractor Aptiv aims to provide their Android platform
with a voice assistant meant to be sold to car manufacturers, but they aspire to
avoid locking their platform to one voice component. This thesis investigates the
possibilities of mitigating the difficulties of integrating multiple voice recognition
components into one unified framework using a design science research approach
over the course of two cycles. Data has been collected through both formal and informal interviews that address
the problem at hand. An Ishikawa diagram was constructed to illustrate an overview
of the problem space at Aptiv and through evaluation, a new narrowed scope could
be determined. The new scope resulted in an investigation of the four voice recognition
components Aptiv showed interest in, which were Amazon Alexa, Google
Assistant, Mycroft, and Houndify. This investigation lead to the creation of a tool
containing 15 parameters developers can use to compare the functionality of different
voice recognition components. The tool was then evaluated by developers at
Aptiv through a workshop. The tool together with its illustrations is a helpful start for developers tasked with
integrating, either one or several, voice recognition components for an Android system.
The presented details of each component are prone to change in the future due
to uncertainties regarding their future development. Further investigation of each
voice recognition system is necessary in order to fully create the general solution
Aptiv aims for. Other companies may benefit from the process that has been used
in this thesis when trying to integrate components. The artifact presented in this
thesis could be applied to other cases, where multiple components are present and
the causes and effects of choosing one or several have implications that need to be
considered.
Beskrivning
Ämne/nyckelord
voice recognition components , automotive software engineering , design science research , developing strategies , interchangeable components , software reuse