New blends of Speech Powered Interactive Voice Response (SEIVR) Systems have replaced the traditional time-consuming menu-driven IVRs. However, when broadcast across telecom networks using narrowband codecs, the voice signal’s quality and intelligibility degrade dramatically. Only a revolutionary technology of Artificial Band Width Extension may provide wideband quality signal without considerable alteration of the existing network infrastructure (ABWE). In this paper, ABWE is implemented with a QMF filter bank that band splits input speech into LF and HF components before compressing and encoding the HF components with a novel data concealing methodology. To build an artificial wideband speech signal, a QMF synthesis filter bank is employed in the reconstruction process. A client-server technique with socket programming on a single machine was utilised to construct the proposed model, assuming no noise and no transmission failures. A comparison analysis was also performed to determine the root cause of SEIVR system performance decrease. In simulations, the suggested SEIVR model with ABWE showed a significant improvement in speech recognition accuracy and overall efficiency. Based on the simulation results, an artificially extended NB speech signal has a recognition accuracy that is greater to that of an NB speech signal and comparable to that of a WB speech signal.
Department of Electronics and Communication Engineering, Sreenidhi Institute of Science and Technology, Hyderabad, Telangana, India and Department of Electronics and Communication Engineering, JNTUH, Hyderabad, Telangana, India.
K. Anitha Sheela
Department of Electronics and Communication Engineering, JNTUH, Hyderabad, Telangana, India.
View Book :- https://stm.bookpi.org/AAER-V6/article/view/1006