Swathi Kotte , pp. 66. ING/School of Engineering, 2011.
Echo cancellation in voice communication is a process of removing the echo to improve the clarity, quality of the voice signals by suppressing the silence signal which prevents echoes during transmission through networks. There are two types of echo in voice communication: Hybrid echo cancellation and Acoustic echo cancellation. Hybrid echo is generated by the reflection of electrical energy by a device called Hybrid. Echo suppressor’s helps to minimize the hybrid echo’s to produce clarity voice signals. The coupling problems between the telephony speaker and its microphone lead to the acoustic echo’s. The direct sound from the loud speaker enters into the microphone almost unaltered this is called direct acoustic path Echo. These echoes may be caused by cross talks or by echo in caller surroundings. These disturbances vary depending on environmental preliminaries such as ventilators, fans, walls and other disturbing sources.
The main objective of this research is to present acoustic echo cancellation design methods. We investigate two parts of echo cancellation design: In the first part we focus on echo cancellation for sinusoidal signal using different algorithms like Least mean square algorithm(LMS),Leaky Least Mean Square (LLMS) Algorithm, Normalized Mean Square (NLMS) Algorithm, and Recursive Least Square(RLS) Algorithm based on different parameters .The second part our work focus on the robustness of Acoustic Echo Canceller(AEC) in the presence of interference with regards to the near end speech theory and implementation aspects for acoustic echo cancellation. This paper presents the comparison between different adaptive filter usages in acoustic echo cancellation. This comparison includes the cancellation of acoustic echo generated in room using different adaptive filter like least mean square (LMS) Algorithm, Leaky Least mean square (LLMS) Algorithm, Normalized Least Mean Square (NLMS) Algorithm and Recursive Least Square (RLS) Algorithm and we also take an input sinusoidal signal and add additive white Gaussian noise and compare this results with the speech signal based on different parameters. We observe the different parameters like Echo Return Loss Enhancement (ERLE), signal to noise ratio, comparing ERLE with different filter parameters, comparing ERLE with filter length and computational complexity. We show a number of experimental results to illustrate the performance of the proposed algorithms and from the results we observe that ERLE value is high for LMS, LLMS algorithm, SNR values is high for NLMS algorithm, different filter parameters are compared with the ERLE and maximum values are estimated. The simulation part is done in MATLAB and the output results are plotted.