Integrated 3D Active Noise Cancellation Simulation and Synthesis Platform Using Tcl

As the virtual reality market has grown, implementing an optimized virtual environment has become more important. Three-dimensional audio rendering is a crucial aspect of virtual reality. External interference, such as acoustic noise, should be minimized to build a system where users can fully immerse themselves in the virtual world. Active noise cancellation allows users to analyze and process the noise in real time. Recently, audio signal processing methods that use deep learning have been widely studied. However, utilizing deep learning models at runtime can be challenging, thus requiring a suitable external simulation environment, including hardware accelerators. In this paper, we propose a Tcl-based active noise cancellation platform for the removal of three-dimensional noise components. This platform constructs a noise library using temporal convolutional networks and selects filter weights to remove various types of noise. The experimental results show that it is possible to construct a library with a minimum delay of 10ms and a size of 36 bytes. This allowed for real time noise cancellation with short delay time and low memory requirement. In addition, various noise environments were experimented with using FPGA, and it was demonstrated that the signal-to-noise ratio improved by an average of 3.8dB.