raymondhuang210129/uiuc-audio-lab

UIUC CS498 Spring 2020: Audio Labaratory Assignments

Introduction

• This course covers the basic ideas of audio processing and wide range of techniques to create sound effects.

• Courses and assignments are designed by Professor Paris Smaragdis.

Contents

• W1: Basic Sound Manipulation

  • Generate tones and basic sounds such as chirp, ping-pong, etc.
  • In/output sounds from/to storage or sound devices

• W2: Time and Frequency Domain

  • Conduct Short Time Fourier Transform and generate spectrograms
  • Blocking some noises in Frequency Domain and transform back to sound data in time domain.

• W3: Fix Sounds with Filters

  • Design Lowpass, Highpass, Bandpass, Bandstop filters with different techniques and apply on different type of noises.

• W4: Reverberation and Room Simulation

  • Generate reverb effects by designing Comb filters, Allpass filters, etc.
  • Create the room's response via deconvolution and apply to the sounds.

• W5: 3D Audio & Virtual Sound

  • Simulate the sound sources by convoluting the original sound with ITD and ILD filters
  • Simulate the dynamic sources by covoluting the sound with different HRTFs at each moment

• W6: Microphone Arrays

  • Generate steering vector for different inputs' phase shifts and identify the directions of sound sources.
  • Enable the microphone array to act like beamforming microphone and tramsform the original input to localized sound.

• W7: Denoising

  • Sample the background noise from recording and denoise with spectral subtraction and median filtering
  • Dynamically sample the noise by using voice activity detector to get better result in every moment

• W8: Pitch Tracking and modifications

  • Track the voice pitch by using autocorrelation and ignore un-pitched frames
  • Track the voice pitch by detecting zero-crossing and tune the sound with pitch synchronous overlap adding (PSOLA)

• W9: Spectral Factorizations

  • Learn the spectral factors in sound track and extract sounds emitted by each instrument
  • Train the factors of human speaking and chimes and separate the mixture of the two

• W10: Audio Classification

  • Train an speech/music classifier with implementing covariance vector and Gaussian PDF function
  • Implement music genres classifier with sklearn API and applying MFCC transformation to enhance the characteristics of each genres

• W11: DTW and HMMs

  • Apply DTW on speech recognizer to recognize user's inputs