Development of a Voice Enhancement Device for Adaptive Multi-Rate Wideband (AMR-WB) Codec: Difference between revisions

From Iskomunidad
(The abstract is from another thesis, this is my thesis so i put its real abstract)
Line 6: Line 6:


      
      
Recent advances in wireless communication schemes have resulted in an increasing demand for small, highly integrated RF transceivers that have very low power dissipation.  Consequently, there has been an aggressive downscaling of RF CMOS technologies, enabling the possibility of denser integration, higher transit frequencies, and lower costs.
Measurements show that the one-way delay in Next-Generation Networking (NGN) mobile communications typically exceeds the limit for acceptable voice quality.  One possible approach to address this issue is to cut down the delay contributed by the suboptimal operation of the Voice Enhancement Device (VED).
      
      
However, CMOS scaling has introduced new design challenges. Complex systems such as wireless receivers could no longer be analyzed simply at the circuit level.  There arises a need to simulate new architectures at a higher level of abstraction so that they can be optimized and compared quantitatively in terms of performance, power and area.  There exists a large dependency between system-level architectural decisions and performance metrics of each circuit block in the architecture, which is in turn dictated by the technology process being used.
This research developed and evaluated a parameter-based VED that operates on coded speech rather than on speech waveforms. The information contained in the speech parameters were utilized for voice enhancement such that no transcoding will be needed.  For acoustic echo control, the approach is to destroy the speech-like characteristics of the echo by modifying the received parameters.  For noise reduction, the traditional procedure is performed except that the approach looks at the codebook gain to estimate the noise level.  This approach for VED can be used in NGN to minimize the speech transmission delay, thus, achieving the acceptable voice-quality.
      
      
In this research, the different system-level issues involved in the design of a direct-conversion receiver are investigated.  The receiver is designed to target Worldwide Interoperability for Microwave Access (WIMAX) applications, which has been rapidly gaining momentum as an alternative to cable and DSL in delivering wireless broadband access to end-users.
For objective evaluation, relevant tests based on ITU-T recommendations, showed that both Acoustic Echo Control (AEC) and Noise Reduction (NR) Functions passed the ITU-T G.160 specifications.  For subjective listening test, NR sections resulted in lower listening-effort mean opinion score (MOSLE). However, improved MOSLE was achieved when both AEC and NR sections were combined.
   
It was found through simulations that the CMOS process introduces large values of flicker noise, which could be a barrier especially for the design of the down conversion mixer and the other baseband clocks such as the filter and the VGA.
   
High-level design trade-offs in the direct-conversion receiver architecture are also investigated to characterize the effect of each block in the whole system’s performance.  Based on system-level simulations, a design flow is developed that would further improve subsequent efforts to design a fully integrated wireless CMOS transceiver, the laboratory.
 




Revision as of 03:08, 18 September 2011

Macario O. Cordel, II

(MS Graduated: 1st Sem 2010-2011)

Abstract


Measurements show that the one-way delay in Next-Generation Networking (NGN) mobile communications typically exceeds the limit for acceptable voice quality. One possible approach to address this issue is to cut down the delay contributed by the suboptimal operation of the Voice Enhancement Device (VED).

This research developed and evaluated a parameter-based VED that operates on coded speech rather than on speech waveforms. The information contained in the speech parameters were utilized for voice enhancement such that no transcoding will be needed. For acoustic echo control, the approach is to destroy the speech-like characteristics of the echo by modifying the received parameters. For noise reduction, the traditional procedure is performed except that the approach looks at the codebook gain to estimate the noise level. This approach for VED can be used in NGN to minimize the speech transmission delay, thus, achieving the acceptable voice-quality.

For objective evaluation, relevant tests based on ITU-T recommendations, showed that both Acoustic Echo Control (AEC) and Noise Reduction (NR) Functions passed the ITU-T G.160 specifications. For subjective listening test, NR sections resulted in lower listening-effort mean opinion score (MOSLE). However, improved MOSLE was achieved when both AEC and NR sections were combined.

Retrieved from "https://iskomunidad.upd.edu.ph/index.php?title=Development_of_a_Voice_Enhancement_Device_for_Adaptive_Multi-Rate_Wideband_(AMR-WB)_Codec&oldid=18530"