| Proj No. | A3055-251 |
| Title | Developing UI and dashboard for Urban sound datasets: Singa:pura datasets |
| Summary | In this innovative project, we embarked on an extensive field operation to record urban sound signals across a diverse array of settings and locations within the urban landscape. The essence of this initiative is to capture the auditory essence of city life, from the bustling streets and crowded public spaces to the more subdued tones of residential areas and parks. This rich collection of in-situ urban sound data serves as a fertile ground for advanced analytical work, aimed at deciphering the complex tapestry of sounds that define urban environments. Students involved in this project are tasked with a comprehensive analysis of the collected sound data. The process begins with meticulous tagging of each sound recording, identifying specific sources and characteristics such as volume, frequency, duration, and temporal patterns. This initial step is critical for structuring the dataset in a way that facilitates further analysis and prepares the data for input into machine learning models. Following the tagging phase, students will delve into signal preprocessing. This involves a series of technical procedures designed to enhance the quality of the sound data, making it more amenable to analysis. Techniques such as noise reduction, normalization, and feature extraction will be employed to isolate relevant audio characteristics and remove extraneous sounds that could interfere with the classification process. The analytical heart of the project lies in developing a deep neural network capable of classifying the various types of urban noise and their characteristics with high accuracy. This involves training the model on the preprocessed datasets, fine-tuning its parameters, and rigorously testing its performance to ensure it can accurately identify and categorize a wide range of urban sounds. The ultimate goal is to create a sophisticated model that can not only distinguish between different types of urban noise but also provide insights into their acoustic properties and implications for urban living. Upon successful development and validation of the model, the plan is to open-source the datasets on DR-NTU, Nanyang Technological University's digital repository. Making these datasets publicly available will significantly contribute to the global research community's efforts in urban sound analysis, potentially leading to breakthroughs in noise mitigation strategies and urban planning. Participation in this project offers students a unique opportunity to develop a robust skill set in big data analysis, specifically in the context of environmental sound research. They will gain hands-on experience with cutting-edge audio processing techniques, machine learning algorithms, and data management practices. Moreover, students will have the chance to contribute to a meaningful project that addresses real-world challenges, preparing them for future careers in data science, urban planning, environmental studies, or any field where big data analysis plays a pivotal role. |
| Supervisor | Prof Gan Woon Seng (Loc:S2 > S2 B2B > S2 B2B 68, Ext: +65 67904538) |
| Co-Supervisor | - |
| RI Co-Supervisor | - |
| Lab | Digital Signal Processing Lab (Loc: S2-B4a-03) |
| Single/Group: | Single |
| Area: | Digital Media Processing and Computer Engineering |
| ISP/RI/SMP/SCP?: |