The demand for voice and video conferencing solutions for remote yet personal communication has inevitably increased significantly in recent years. The added value of these systems will continue to be utilised beyond the pandemic period. BigBlueButton (BBB) is one such open-source web conferencing system that is the first choice at the University of Oldenburg. For natural communication, it is necessary, among other things, to prevent acoustic crosstalk between the microphone and loudspeaker, which is perceived as an annoying echo of one's own voice. BBB uses automatic acoustic echo control (AEC) for this purpose. This AEC is part of the audio processing module, which works together with other algorithms to improve the microphone signal. The module is based on the WebRTC standard, which is one of the most common web browser libraries and enables real-time communication via peer-to-peer connections.

The bachelor thesis will provide an overview of the processing modules of the BBB web conferencing service, in particular WebRTC. The link between the modules and their interfaces to the outside world will be worked out, focussing on the AEC of the
audio processing module. The source code provided by WebRTC/BBB provides the basis for this. In order to demonstrate the structure that has been worked out, the processing and transmission of the signals processed by the BBB audio processing module will be analysed. Speech signals selected from relevant databases are used for the analysis. The microphone signals to be processed and the processed transmission signals from the BBB module are to be evaluated in the bachelor thesis using relevant instrumental metrics, such as ERLE in individual speech, PESQ in intercom speech or AECMOS in the overall context.

Contact: Prof. Dr.-Ing. Gerald Enzner room W2-0-068) and Svantje Voit room W2-074).

The BENCH algorithm (Blind Equalisation and Channel Identification) is a method for automatically determining the useful signal direction of a microphone array. This can be a linear, circular or arbitrary arrangement of microphones, as the acoustic transfer function from the source to the microphones is regarded as abstract directional information. BENCH also uses this information to focus on the useful signal in order to improve the noise and reverberation characteristics of the recording room. The formulation of the BENCH algorithm and the proof of its function was provided for concentrated microphone arrays.

Its task is to design a distributed microphone array as it can be foreseen in the scenario of ad-hoc formed acoustic sensor networks. The design should be based on the idea of several video conference participants with several devices in the same room. This situation currently harbours considerable technical problems in practical operation and therefore poses challenges for research. In the thesis, the BENCH algorithm is to be executed on the relevant microphone signals and evaluated with regard to speech quality and noise filtering. Commercially available multi-channel sound cards and Matlab/Simulink software for signal processing on PCs are used as tools for the real-time execution of the algorithm.

Contact: Prof Dr Gerald Enzner Room W2-0-068) and Dr Aleksej Chinaev Room W2-073)

Algorithm research in the field of speech and audio signal processing likes to use powerful and practical PC-based platforms for prototypical implementation with real-time behaviour. Signals are acquired, processed and prepared for demonstrative playback with various listening options. Various driver and signal processing architectures are available for both Windows and Linux operating systems. There is also a need to transmit the signals to a listening room or to a remote end, such as in the context of voice communication.

In the bachelor thesis, a PC-based system for audio recording and algorithmic audio signal processing is to be set up. Depending on experience, either a Linux or Windows operating system will be used. The components of the system are arranged from analogue microphone recording, analogue amplification and digitisation through to acquisition in processing software (such as Simulink). A particular focus of the work concerns the transfer of processed signals (or unprocessed signals in the simplified system) to a communication solution such as Zoom or Big Blue Button. In this context, the effects of "virtual audio cables" for connecting software components for signal processing and signal transmission are analysed.

Contact: Prof Dr Gerald Enzner room W2-0-068) and Dr Aleksej Chinaev room W2-073)