|
Title:
|
AI in Cyber Security & System Solutions
|
|
Author:
|
Bulkesh Sinhmar, Dr Bijender Bansal, Dr Vinit Kumar Lohan and Dr Deepak Goyal
|
|
Keywords:
|
Multimodal Deep Learning, Skeletal Motion Analysis, Cross-Cultural Dance Recognition and Cultural Heritage Informatics and Explainable AI (XAI).
|
|
Abstract:
|
The groundwork for event detection in football match analysis using audio. It describes the structure, annotated event categories and preparation of synchronized audio data for the Soccer Action Compilation (SADC) dataset. The chapter outlines the major steps in the pre-processing such as segmentation, noise reduction and feature extraction of match-sounds using Mel-Frequency Cepstral Coefficients (MFCCs), which are important for the temporal and spectral properties. The design and implementation of the audio event detection model are explained and the key event detection such as goal, goal attempt, penalty and yellow card is explained with the use of deep learning architectures. Training configurations, parameter selection and evaluation strategies are discussed to make sure that the models are reliable and reproducible. Results of the model are presented using quantitative measures and confusion matrices, and are followed by a discussion on important results, strengths and limitations that were observed. In general, this chapter provides the methodological and analytical framework for the subsequent visual and multimodal event detection in football videos, while showing the effectiveness of using audio cues in understanding the dynamics of a game and summarising the key moments.
|
|
Download Paper:
|
|
|