Multi-Lingual Speaker-Independent Voice User Interface For Mobile Devices

Optimizing speech naturalness in voice user interface design: A weakly-supervised approach

“…Hence, choosing a design tool that will enhance portability and product robustness comes to question. Also issues concerning product adaptation and multilingualism [2] must be carefully resolved. In addition, users expect an application that is simple, flexible and less demanding.…”

“…Speech technology (speech synthesis-TTS and speech recognition) permeates several interactive-driven devices. Its techniques pervade several fields and applications such as telecommunication services, most of which have embraced TTS to convey information about timetables and events; measurements and control systems, as exemplified in the speaker-independent name dialling system found in many cell phones [2][3][4][5]; optical character recognition for the physically challenged [6][7]; talking systems for 'hostile' environments; books and toys; etc. We incorporate higher-level linguistic analysis (syntax and morphology) in a weakly-supervised manner using the Hidden Markov Model (HMM) [8] to train an enriched front-end of an African tone language system.…”

Speech‐Recognition Overview

“…It is usually combined with CMN and other techniques, such as speaker adaptation, and yields good results. Some embedded speech recognition systems [78] use CMN and CVN because of their low implementation costs. CMN and CVN are further extended in [79] to equalize histograms of training and testing data.…”