Publication detail

Optimization of Mel-Frequency Cepstral Coefficients for Automatic Gunshot Detection

HRABINA, M.

Original Title

Optimization of Mel-Frequency Cepstral Coefficients for Automatic Gunshot Detection

Type

conference paper

Language

English

Original Abstract

Mel-frequency cepstral coefficients (MFCCs) are features widely used in tasks such as sound event detection (SED), speech recognition or other audio processing problems. This paper explores various ways of modification MFCCs in order to achieve optimal detection results for gunshot detection task. The paper focuses especially on frequency scale and bandwidth, number of filter banks and filter bank shape. The final results achieve recognition performance of 98.7 true negative rate (TNR) and 84.8 true positive rate (TPR).

Keywords

gunshot detection, mel-frequency cepstral coefficients, optimization

Authors

HRABINA, M.

Released

10. 9. 2019

ISBN

978-80-214-5781-2

Book

Proceedings of IEEE Student Branch Conference Mikulov 2019

Pages from

46

Pages to

48

Pages count

3

BibTex

@inproceedings{BUT159769,
  author="Martin {Hrabina}",
  title="Optimization of Mel-Frequency Cepstral Coefficients for Automatic Gunshot Detection",
  booktitle="Proceedings of IEEE Student Branch Conference Mikulov 2019",
  year="2019",
  pages="46--48",
  isbn="978-80-214-5781-2"
}