Aalto University School of Electrical Engineering,
Department of Signal Processing and Acoustics,
2017
Advisor: D.Sc. Kalle Palomäki
Understanding and participating in conversations has been reported as one of the biggest challenges hearing impaired people face in their daily lives. These communication problems have been shown to have wide-ranging negative consequences, affecting their quality of life and the opportunities available to them in education and employment.
A conversational assistance application was investigated to alleviate these problems. The application uses automatic speech recognition technology to provide real-time speech-to-text transcriptions to the user, with the goal of helping deaf and hard of hearing persons in conversational situations. To validate the method and investigate its usefulness, a prototype application was developed for testing purposes using open-source software. A user test was designed and performed with test participants representing the target user group.
The results indicate that the Conversation Assistant method is valid, meaning it can help the hearing impaired to follow and participate in conversational situations. Speech recognition accuracy, especially in noisy environments, was identified as the primary target for further development for increased usefulness of the application. Conversely, recognition speed was deemed to be sufficient and already surpass the transcription speed of human transcribers.