School of Engineering Department of Electronic and Computer Engineering 138 Department of Electronic and Computer Engineering Projects in Audio Signal Processing Supervisor: CHAU Kevin / ECE Student: GUO Shiran / CPEG Course: UROP 1000, Summer With the development of audio modification techniques, various readily available free software can now perform audio splicing. However, splicing forgery on audio segments can affect the authenticity of the materials for forensic detection. This report will review and summarise the techniques implemented for audio splicing forgery detection presented in various academic papers. It will start with the intuitive idea of directly comparing the spectra of different approaches to make the algorithms robust to several commonly used post-processing operations, followed by the more computationally efficient approaches. Methods for heterogeneous splice detection will then be presented. Finally, we will summarise and propose prospects for future development. Projects in Audio Signal Processing Supervisor: CHAU Kevin / ECE Student: HE Weike / ELEC Course: UROP 1100, Summer Presently reported is automatic surtitle generation scheme for live Cantonese opera based on beat tracking. With an impulse train prepared from a music-vocal score of the opera, the individual Chinese characters in the surtitles can be animated in a karaoke style in sync with the performance automatically. Specifically, the impulse train registers the onset position of every monosyllabic Chinese character to be sung or spoken. The impulse train also synchronizes with the live performance by monitoring the beats in the streaming audio. The major advantage of this method is the ability to predictively highlight the characters at the onsets of vocalization with almost zero latency. Assuming beat tracking is accurate, the timing precisions of the surtitle characters in singing parts should be 100% correct. Projects in Audio Signal Processing Supervisor: CHAU Kevin / ECE Student: JIA Yusheng / COMP Course: UROP 1000, Summer This semester’s learning process gave me a glimpse of audio signal processing methodologies that are specific to music and used in applications. Some basic measures were studied for the theoretical parts, including the Discrete Fourier Transform and the Short Time Fourier Transform. In addition, some analysis models, such as sinusoidal, harmonic, and stochastic models, are briefly introduced. The learning focused on spectral processing techniques of practical knowledge with which to analyse, synthesise, transform, and describe audio signals in music applications. In this report, I will mainly discuss my understanding of these fundamental theories, my harvest of scientific research, and the transformation from task-driven learning to self-motivated learning.
RkJQdWJsaXNoZXIy NDk5Njg=