Abstract
Proposed new wireless communication systems such as third generation cellular and PCN will utilize speech interpolation, disconnecting the user from the spectral resource during pauses in speech in order to reduce radiated emissions and improve spectral efficiency. An accurate model of the on-off characteristics of conversational speech is thus necessary to analyze system performance, particularly if the system utilizes a time and/or frequency division multiple access technique. Previously developed speech activity models are deficient because they either do not reproduce short silent pauses of less than 200 ms. (representative of the silence gaps between syllables or words) or else they do not replicate the dynamics between the two conversing parties. Starting with the P.T. Brady model and developing appropriate modifications, this paper formulates a simple, accurate, comprehensive 8-state Markov model for voice activity in conversational speech. The new model can easily be incorporated into simulations or analyses assessing the performance of various new-generation wireless networks, thus improving the accuracy of the performance assessments.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call