ELEC-E5500 - Speech Processing, Lecture, 8.9.2022-17.10.2022
Kurssiasetusten perusteella kurssi on päättynyt 17.10.2022 Etsi kursseja: ELEC-E5500
Osion kuvaus
-
-
The goal of this exercise is to get acquainted with basics of speech processing. This includes recording, reading, resampling, windowing, and computing magnitude spectrum and spectrogram along with visualizations.
-
In this exercise, you will be implementing two popular fundamental frequency estimation methods: They are: (1) auto-correlation and (2) cepstral methods.
Please read carefully, the instructions provided in the Jupyter notebook. -
Exercise 1: Solution Kansio
-
In this exercise, you will be implementing functions to extract simple speech features which are suitable for voice activity detection (VAD) and utilize them to train two simple VAD classifiers. The computed features are combined into a matrix that acts as the input data provider for our classifiers. The code for the classifier training and some features are already provided in the notebook, but you must experiment with different features and properties, and report your findings.
-
Exercise 2: Solution Kansio
-
The goal of this exercise is to implement basic speech enhancement techniques and evaluate and visualize the quality of the enhancement. You will be implementing four different methods: Spectral-subtraction, Wiener-filter, linear-filter and a VAD based filter.
-
Exercise 3: Solution Kansio
-
Exercise 4: Solution Kansio
-