A New Technique for Artificial Bandwidth Extension of Speech Signal and its Performance Analysis
Tejal Chauhan1, Shraddha Singh2, Sameena Zafar3
1Tejal Chauhan, Research Student, E&C, Patel College of Science and Technology, Bhopal, India.
2Shraddha Singh, Assistant Professor, E&C, Patel College of Science and Technology, Bhopal, India.
3Sameena Zafar, Assistant Professor, E&C, Patel College of Science and Technology, Bhopal, India.
Manuscript received on March 22, 2013. | Revised Manuscript received on April 12, 2013. | Manuscript published on April 30, 2013. | PP: 342-347 | Volume-2, Issue-4, April 2013. | Retrieval Number: D1440042413/2013©BEIESP

Open Access | Ethics and Policies | Cite
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: In current scenario of wireless communication system, quality of voice output is degraded due to its limited bandwidth (300-3400 Hz) and power constraints which in turn offers speech sounding muffled and thin. Recent wireless systems involved in transmission of speech demands evolution of efficient and effective methods for maintaining quality of speech, especially at the receiving end. In order to obtain toll quality of speech and high intelligibility cum naturalness in wireless systems, NB speech coders should be upgraded to its counterpart WB coders (50-7000Hz). For the effective utilization of WB speech communication in wireless media, it is indeed necessary to upgrade both end devices and network to be WB compatible which is costly and time consuming affairs. In the meantime some techniques have been developed to artificially extend bandwidth of NB speech to WB at receiver which leads to improvement in the quality of recovered speech. Amongst all elements of the communication system (channel, transmitter and receiver), quality and intelligibility of voice at receiver side majorly depend upon channel condition. Many techniques are adopted to mitigate the effect of the channel. In order to maintain quality and naturalness of voice at receiver side in various unpredictable channel conditions, AMR (Adaptive Multi Rate) NB is considered to be one of the obvious potential candidates. AMR NB is operated on various modes of bitrate between 4.75 and 12.2 kbps. Depending upon the channel conditions, specific mode of operation is selected dynamically. For example, Low bit rate mode of operation is selected in bed channel conditions, that allows more error protection bits for channel coding and vice versa. Since inception, many speech coding techniques like CELP, ACELP and RPE-LTP are adopted in different applications in 2G and 3G. In this paper, implementation of ABWE algorithm is developed on CELP based GSM AMR 06.90 NB Coder using MATLAB simulation; further Subjective (MOS) and Objective (PESQ) analysis are carried out to judge the overall performance of developed coder. The evaluated results for both analyses clearly advocate that BWE coder offers significant improvement in recovered speech quality in comparison with legacy GSM AMR NB decoder.
Keywords: ABWE, AMR, CELP, GSM, Speech coding, Steganography, Subjective Analysis, Objective Analysis.