A Method Of Automatic Updating Noise Threshold In Double - Ended Call Detection Algorithm

With the rapid development of science and technology, a wide variety of communication methods facilitate communication between people, the rapidly growing popularity of the hands-free phone and video conferencing systems, the people liberated hands, talk more freely, people from the communications equipment has become farther away, the distance to the pickups and speakers do not have to be limited to the mouth and ears. However, this improvement also brought had to increase the speaker volume and pickups sensitivity, it was found that the acoustic echo cancellation has become increasingly obvious bottleneck, if not properly handle the echo problem will result in a communication system double talk worse, severe choppy audio, unable to complete the task of more individuals talk at the same time. Actual the theoretical acoustic echo cancellation technology research constantly applied to a wide variety of products, also encountered a lot of specific issues, such as noise in the actual use of the changing size and double talk detection performance. There are many theoretical issues, how the research means in reality practical problems need people to continue to improve.Adaptive echo cancellation technology and double talk detection technology is used in acoustic echo cancellation technology. The two technologies complement each other to achieve a complete echo cancellation. For acoustic echo cancellation theory better able to practical changing the problem of noise encountered in practical applications, the research work of the following aspects:First echo cancellation history, the current structure of the adaptive filter of the mainstream, filtering principle and adaptive filtering algorithm. An adaptive filtering algorithm has been a variety of implementation methods, and can basically meet the needs of practical application. Due to their different characteristics, respectively, for a variety of special scene, the article focuses on the various statistical indicators of adaptive filtering algorithms, and these indicators in the differences between the various algorithms and the contradictory relationship between them, as well as how to choose the most suitable for the specific application of the actual use of adaptive filtering algorithms.Then further introduced several double-ended talk detection methods. The purpose of the introduction of the double-ended talk detection algorithm in order to inform adaptive filtering algorithm double talk state when it is necessary to suspend the work, to avoid double talk state continue to work and cause filter divergence adaptive filtering algorithm. Call the theory of the double-ended is very detailed, but the theory will encounter many problems in practical applications, such as the most typical is the actual use of the environment, not only to consider the noise brought many theoretical algorithm reasoning directly omitted the problem, but also to consider the noise in a real environment, the range of additional problems brought about by the change.Then, for the double-end call detect input signal SNR uncertainty in the actual environment, noise share percentage will change in the actual environment and other issues, this article was proposed and discussed further. In the double talk detection algorithm, proposed a way to automatically adapt to changes in the noise threshold, and gives a specific calculation method. It can effectively improve the detection sensitivity of double-end call, making the system unstable SNR environment is also able to play a good performance, i.e. in a certain probability of false alarm, as much as possible to reduce the leakage alarm probability.Further, by the voice recognition detection algorithm and the correlation detection algorithm, improving the recognition rate of the system in conditions of low SNR noise so that the system can easily identify the status of only the noise of the local sound, thereby performing the noise level profile, the new duplex detection algorithm by constantly update the noise value to accurately determine the duplex state, and thus control the coefficients of the adaptive filter convergence process.Finally, the computer simulation is given. The simulation results proved that the noise adaptive estimation accuracy, based on the comparison of the cross-correlation detection algorithm to verify the effectiveness of adaptive noise threshold algorithm validated based on the mainstream algorithms improved performance and conclusions are given.