Audio Processing Laboratory, Winter Term 2022/23

image1 image2 image3 image4
  • Instructor: Prof. Dr. Nils Peters
  • Organization: Lorenz Schmidt
  • Format: Virtual Laboratory, in English
  • Credits: 2.5 ECTS
  • Location: Via Zoom. Link and access information for Zoom meetings via StudOn (see below).
  • Dates: Introduction session: 18.10.2022, 11:00

Objectives

The objective of this lab course is to give students a hands-on experience in audio processing. In particular, functions, transforms, and algorithms that are important for analyzing and processing audio signals are covered. The lab course is supervised by AudioLabs staff members.

Requirements

  • Must have experience with Python and NumPy
  • Solid mathematical background
  • Good understanding of fundamentals in digital signal processing
  • General background and personal interest in audio

Experience with the R Programming language is beneficial (beginners tutorial is provided in the course material).

This Audio Processing Laboratory is fully virtual. All participating students must have access to a computer equipped with:

Enrollment

Mandatory enrollment via:

logo-studon

Enrollment phase: Sep 26 2022 - Oct 2 2022

The number of seats is limited! first come, first served.

Schedule

  • The Audio Processing Laboratory consists of one introduction session and five different lab units (see below)
  • In the introduction session, students will be grouped in teams of 2-3 students and select four mandatory lab units
  • Students work remotely on each lab units on their own schedule in groups of 2-3 students
  • Questions can be asked via email and via the StudOn forum
  • In the weeks of the lab units there are dedicated open question sessions on Mon 11:15 - 12:15
  • Some homework exercises are due in handwritten form on paper. These must be submitted on Tuesdays, 12:00 via StudOn (as scans or photographs)
  • Student groups will be examined via video conferencing, including screen sharing. All solutions must be ready for the examination. These examinations will take place on Thursdays 12:00 - 16:00 and Fridays 10:00 - 14:00. Student groups will be assigned a slot among these times
Introduction session
  • Date: 18.10.2022, 11:00 via Zoom (link via StudOn)
  • All prospective students must attend this session to be eligible for the following lab units
Lab 1: Short-Time Fourier Transform and Chroma Features
  • About: This lab unit introduces the Short-Time Fourier Transform (STFT). Students learn how to compute and interprete the discrete STFT and how to visualize its magnitude in a spectrogram representation. From the STFT, students derive various audio features (e.g., Chroma features) which are useful for analyzing music signals.
  • Supervisors: Yigitcan Özer, Simon Schwär
  • Question session: 07.11.2022, Mon, 11:15 - 12:15
  • Homework submission: 08.11.2022, Tue, 12:00
  • Lab Exams:
    • 10.11.2022, Thu, 12:00 - 16:00
    • 11.11.2022, Fri, 10:00 - 14:00
Lab 2: Statistical Methods for Audio Experiments
  • About: Subjective listening tests are a crucial part of assessing and improving the quality of audio technology, such as perceptual audio compression or loudspeaker design. This unit will teach students the basics of experimental statistics as they are used for evaluating auditory experiments. Common testing methodologies are introduced, and students will learn how to use statistical tools to analyze listening test responses.
  • Supervisors: Pablo Delgado, Martin Müller
  • Question session: 28.11.2022, Mon, 11:15 - 12:15
  • Homework submission: 29.11.2022, Tue, 12:00
  • Lab Exams:
    • 01.12.2022, Thu, 12:00 - 16:00
    • 02.12.2022, Fri, 10:00 - 14:00
Lab 3: Speech Analysis
  • About: Speech is the primary means of human communication. Understanding the main properties of speech enables us to develop efficient algorithms for important communication tools, such as mobile phones. The purpose of this unit is to demonstrate challenges in analyzing the most basic and important properties of speech signals.
  • Supervisors: Ning Guo, Richard Füg
  • Question session: 12.12.2022, Mon, 11:15 - 12:15
  • Homework submission: 13.12.2022, Tue, 12:00
  • Lab Exams:
    • 15.12.2022, Thu, 12:00 - 16:00
    • 16.12.2022, Fri, 10:00 - 14:00
Lab 4: Speech Enhancement Using Microphone Arrays
  • About: This unit is designed to give a practical understanding of performing speech enhancement using microphone arrays. Students will implement the spatial beamforming signal processing technique, and analyze the performance of two different beamformers, a signal-independent beamformer and a signal-dependent beamformer. The beamformer performances are compared via objective measures to demonstrate their advantages.
  • Supervisors: Julian Wechsler, Carlotta Anemüller
  • Question session: 16.01.2023, Mon, 11:15 - 12:15
  • Homework submission: 17.01.2023, Tue, 12:00
  • Lab Exams:
    • 19.01.2023, Thu, 12:00 - 16:00
    • 20.01.2023, Fri, 10:00 - 14:00
Lab 5: Convolution and Correlation for Real-time Audio Processing
  • About: In this unit you will learn how to efficiently realize long FIR filters using fast convolution techniques. These techniques are essential for many modern real-time DSP applications. You will examine, program, and evaluate implementations in the time domain and the frequency domain. Because there is a strong relationship between convolution and correlation, you will also learn to implement correlation algorithms e.g., for the estimation of time delays between two signals.
  • Supervisors: Lorenz Schmidt, TBD
  • Question session: 30.01.2023, Mon, 11:15 - 12:15
  • Homework submission: 31.01.2023, Tue, 12:00
  • Lab Exams:
    • 02.02.2023, Thu, 12:00 - 16:00
    • 03.02.2023, Fri, 10:00 - 14:00

Assessment criteria

  • Individual points for each student will be given by the lab examiners (Points: 0=no pass, 1=minimal pass, 2=pass, 3=excellent).
  • To pass the course, all following criteria must be fulfilled:
    • Must attend all four pre-selected units
    • At least one point in all four pre-selected programming units
    • At least six points in total

Links