Abstract:
Due to increase in taxi hailing service (Uber & Careem), a demand for Urdu
text to speech system was increasing. This system would be used in a Pakistani navigation app which would facilitate the drivers(captains) in finding
a way without keeping an eye on the mobile. The goal is to provide Pakistani people the language in which they are most comfortable i.e. Urdu.
In Pakistan, most of the drivers are illiterate and don’t understand English
commands by Google Maps. So we developed an Urdu Text To Speech Sys tem in which the commands during turn by turn navigation would be in
Urdu. Some of the main steps involved are Lexicon generation, LTS rules
generation, Data labeling, Pitch marking, extracting Mel Frequency Cep stral Coefficents, building HMM model. Once the model is generated, we
will export it to Android devices. We will check the quality of both models
i.e. original and the one ported to Android. Our testing is done using both
manual and automatic methods.
