Adaptive multi-resolution analysis of audio signals

Presenter: Nicolas Figueiredo

Spectral representations of temporal signals are vastly utilized in several areas including engineering, mathematics and computer science. For this, the SFTF is commonly used, and it utilizes equally spaced constant resolution frequency components. In order to minimize the trade-off between time and frequency resolution, multi-resolution techniques utilize frequency components with variable resolution and spacing. This same trade-off is the motivation behind adaptive transforms, which utilize information about the signal being analyzed to compose a spectrogram that prioritizes frequency or time resolution depending on the signal's characteristics.

In this seminar, the Constant-Q Transform (CQT) will be presented as an example of a multi-resolution analysis, along with two types of adaptive transforms developed by Lukin & Todd and the idea behind a possible contribution based on reducing the computing costs of adaptive transforms.</p><p>

(video presentation in portuguese)
</p>

When: April 10th, 2019

Where: Auditório do CCSL/IME