Research And Implementation Of Audio Media Engine Based On WebRTC

With the rapid development of computer networks and the continuous advancement of society,the distance in space is no longer an obstacle to communication.People in different regions can easily communicate with other people using audio and video applications.These audio and video applications provide convenient and cheap communication methods that have revolutionized the communications industry.Most audio and video applications rely on multimedia processing engines,and the audio engine and the audio processing algorithms it contains are an important part of it.It is very difficult for ordinary developers to implement a set of audio engines independently.It is necessary to combine audio media capture,audio encoding and decoding,audio processing and real-time transmission control.We can rely on Google Open Source Web RTC(Web Real-Time Communications)Technology to achieve it.In order to make better use of the related technologies in Web RTC and realize the complete audio engine function,this thesis deeply studies the audio engine of Web RTC,and studies the audio transmission,coding and processing algorithms it involved.The principle of the audio processing algorithm has been deeply analyzed and studied,and the actual performance of the audio processing algorithm was tested.For the purpose of improving the performance of the noise suppression algorithm,we propose and implement a deep neural network(DNN)based noise suppression algorithm for common noise in actual audio and video conference systems.The experimental results show that the DNN-based noise suppression algorithm has better performance than the Web RTC noise suppression algorithm for white noise,fan noise,air conditioning noise,airflow noise and crowd noise under low SNR conditions.Especially for crowd noisy noise,DNN-based noise suppression algorithm has a clear lead in many indicators.Finally,we implement the full functionality of Web RTC's audio engine and tested the capabilities of the audio engine.The full text of this article is planned as follows: The first chapter is the introduction part,introducing the research background,the overall architecture of Web RTC and the overall architecture of the Web RTC audio engine.The second chapter studies the audio codec and transmission part of the audio Web RTC audio engine,focusing on the audio codec i SAC and i LBC recommended by Web RTC,the Net EQ algorithm,and the RTP and RTCP protocols for real-time multimedia data transmission.The third chapter studies the audio processing module in the Web RTC audio engine,focusing on echo cancellation,noise suppression and voice activity detection,and experimentally tests the performance of the algorithm.The fourth chapter introduces the algorithm principle of the proposed DNN-based audio denoising algorithm,implements the algorithm,and evaluates the noise suppression effect of the algorithm objectively in detail.It also compares with the performance of Web RTC's noise suppression algorithm.The fifth chapter is the implementation and performance evaluation of Web RTC audio engine and the complete implementation and testing of Web RTC audio and video system.The sixth chapter is the summary and outlook.