Summary of Research Activities: Attending this conference was a fruitful experience. I would like to highlight two things gained at the conference. The rst thing was about sound systems, which was a major topic at the conference. Although digital systems predominated the consumer market, analogue systems were still used extensively when ultra-low latency was required. One use-case of analog sound systems was in opera houses. When a singer sang on the stage, the singer also heard the voice using an earphone. If a digital sound capturing device was used, the digital-to-analog and analog-to-digital conversion might result in delay in the playback of the sound heard by the singer. To minimize the delay, analogue systems were used to avoid conversions between analog signals and digital signals. The second thing was about the sound classi cation systems using deep learning, which was another major topic at the conference. With the success in machine learning, machine learning models had been widely adopted, but human found it dif cult to understand what the model had learnt. A researcher presented an interpretable deep learning model for automatic sound classi cation, which included a prototype layer to assist humans in understanding the workings inside the model, thus helping humans to predict the model’s behaviour. The use-case of such design might not be limited to sound classi cation. HE Weike Major : ELEC Supervised by : Prof. CHAU Kevin / ECE Conference : 156th Convention of the Audio Engineering Society Venue : Madrid, Spain Duration : June 15–17, 2024 ix
RkJQdWJsaXNoZXIy NDk5Njg=