/Audio_Signal_Processing_For_Music_Applications_idu

Analysis of public audio datasets and audio datasets of my making

Primary LanguageJupyter Notebook

Introduction

The codes include audio analyses of coughing and sneezing wave sounds, recorded speech of me pronouncing different letters, and other publicly available annotated and standard audio datasets.

The Folders:

'Feature_Exctraction' folder

  • The code checks wave sound files of two different categories, cough and sneez. Cough and sneez files exist in the 'data' folder above.
  • The code works on features extraction from the chosen wave sounds. The features extracted from the wave files include the amplitude frequency, spectrogram, Mel-frequency cepstral coeffecients and energy band ratio.
  • The code includes results-clearfying comments.

'Audio_Analysis_on_my_voice' folder

  • The codes use my voice as data. Three wave sounds were recorded of me saying three different letters 'a', 'e', and 'o'. The voice files are in 'data' folder, 'kenansounds'.
  • The code implements Windowing, Formants, and Convertion to midi chromagram to all three wave sounds.
  • Comments on the results are in the code.

'Pitch Exctraction' folder

  • The code uses the standard 'orchset' dataset. The dataset can be downloaded from https://zenodo.org/record/1289786#.YKqUs6IzZH5.
  • The code collects ground truth values from the satndard data.
  • The code makes predictions using Essentia-Melodia and Crepe Algorithms and check the prediction results in terms of Voicing Recall, Voicing False Alarm, Raw Pitch Accuracy, Raw Chroma Accuracy, and over all accuracy.
  • Comments on the performance of the two algorithms are at the end.