Each module leads you through the material in a specific order: simply follow the tabs from left to right. But do also explore the entire course content on your own, look ahead to what is coming later, and browse the rest of this website including the forums.
This course operates a policy of continuous improvement. You will find some content is being revised for 2023-24. Don’t worry, it will be ready in time for you to study it. If you looked at previous versions of the course, the main difference is that we are introducing online quizzes through the semester rather than having one big exam at the end of semester. This was in response to the feedback that students didn’t get feedback, especially on the phon and signals component, until very late in semester. Also, hopefully spreading this out will relieve a bit of the exam/final paper stress that happens in early December.
Unlike last year, we won’t be having a separate Q&A slot as these were not particularly well attended last year. You should still feel free to ask questions during lectures, labs, and on the speech.zone forum!
The readings are specified in each module and are categorised as
- Essential (read all, aiming to complete within the week after the lecture on the module topic)
- Recommended (read if you have time, these readings will deepen your understanding but are not essential)
- Extra (only read if you’re interested or looking for extension material; some readings may be challenging; most are beyond the scope of the course)
- Friday, 22 September 2023
-
-
Milestone - enrolFriday, 22 September 2023Enrol for the course on EUCLID.
-
- Wednesday, 27 September 2023
-
-
CANCELLED: Speech Processing Lab - Group 1Wednesday, 27 September 2023 at 09:00 - 11:00Week 2 labs are cancelled due to the UCU strike
Praat basics and phonetics exercises
Before this lab: read Module 1 - Phonetics and Representations of Speech - Lab
Staff: Catherine Lai, Rebekka Puderbaugh -
CANCELLED: Speech Processing Lab - Group 2Wednesday, 27 September 2023 at 16:00 - 18:00Week 2 labs are cancelled due to the UCU strike
Praat basics and phonetics exercises
Before this lab: read Module 1 - Phonetics and Representations of Speech - Lab
Staff: Catherine Lai, Rebekka Puderbaugh
-
- Thursday, 28 September 2023
-
-
CANCELLED: Lecture 2: Acoustic phoneticsThursday, 28 September 2023 at 09:00 - 11:007GS_F.21, 7 George SquareWeek 2 lecture has been cancelled due to the UCU strike
Before this class: complete Module 2
Lecturer: Rebekka Puderbaugh
-
- Wednesday, 04 October 2023
-
-
Speech Processing Lab - Group 1Wednesday, 04 October 2023 at 09:00 - 11:00Acoustic phonetics and speech segmentation
Before this lab: read Module 2 - Acoustics of Consonants and Vowels - Lab
Staff: Rebekka Puderbaugh -
Speech Processing Lab - Group 2Wednesday, 04 October 2023 at 16:00 - 18:00Acoustic phonetics and speech segmentation
Before this lab: read Module 2 - Acoustics of Consonants and Vowels - Lab
Staff: Rebekka Puderbaugh
-
- Thursday, 05 October 2023
-
-
Lecture 3: Digital Speech SignalsThursday, 05 October 2023 at 09:00 - 11:007GS_F.21, 7 George SquareWhat are spectrograms really? An introduction to Digital Signal Processing and the Discrete Fourier Transform
Before this class: complete Module 3
Lecturer: Catherine Lai
-
- Wednesday, 11 October 2023
-
-
Speech Processing Lab - Group 1Wednesday, 11 October 2023 at 09:00 - 11:00Interpreting the DFT and issues with digital signals (Jupyter Notebooks)
Before this lab: read Module 3 - Digital Speech Signals - Lab
Staff: Atli Sigurgeirsson, Catherine Lai -
Speech Processing Lab - Group 2Wednesday, 11 October 2023 at 16:00 - 18:00Interpreting the DFT and issues with digital signals (Jupyter Notebooks)
Before this lab: read Module 3 - Digital Speech Signals - Lab
Staff: Zihang Peng, Catherine Lai
-
- Thursday, 12 October 2023
-
-
Lecture 4: The Source-Filter ModelThursday, 12 October 2023 at 09:00 - 11:007GS_F.21, 7 George SquareBuilding on our understanding of the digital signal processing, we look at source-filter model from more of an engineering perspective
Before this class: complete Module 4
Lecturer: Catherine Lai
-
- Monday, 16 October 2023
-
-
Assessment - Phon/Signals online quiz (10% of course mark)Due on Wednesday, 18 October 2023 by 12:00Multiple choice questions: Test open for 2 days, but once you start you need to complete within 1 hour on Learn. This test covers modules 1-3.
-
- Tuesday, 17 October 2023
-
-
Assessment - Phon/Signals online quiz (10% of course mark)Due on Wednesday, 18 October 2023 by 12:00Multiple choice questions: Test open for 2 days, but once you start you need to complete within 1 hour on Learn. This test covers modules 1-3.
-
- Wednesday, 18 October 2023
-
-
Speech Processing Lab - Group 1Wednesday, 18 October 2023 at 09:00 - 11:00Exploring the source-filter model (Jupyter notebooks).
Before this lab: read Module 4 - the Source-Filter Model - Lab
Staff: Atli Sigurgeirsson, Catherine Lai
(If there is time left over, you can start playing with Festival.) -
Assessment - Phon/Signals online quiz (10% of course mark)Due on Wednesday, 18 October 2023 by 12:00Multiple choice questions: Test open for 2 days, but once you start you need to complete within 1 hour on Learn. This test covers modules 1-3.
-
Speech Processing Lab - Group 2Wednesday, 18 October 2023 at 16:00 - 18:00Exploring the source-filter model (Jupyter notebooks).
Before this lab: read Module 4 - the Source-Filter Model - Lab
Staff: Zihang Peng, Catherine Lai
(If there is time left over, you can start playing with Festival.)
-
- Thursday, 19 October 2023
-
-
Lecture 5: Speech Synthesis - Phonemes and the Front-EndThursday, 19 October 2023 at 09:00 - 11:007GS_F.21, 7 George SquarePronunciation, including letter-to-sound models, and predicting prosody. All these tasks can be done with Classification And Regression Trees (CARTs).
Before this class: complete Module 5
Lecturer: Catherine Lai
-
- Friday, 20 October 2023
-
-
Milestone - assignment 1Friday, 20 October 2023Prepare for the first assignment by reading through the complete instructions on speech.zone. Start work on the assignment, using the forums to get any technical help you need.
-
- Wednesday, 25 October 2023
-
-
Milestone - assignment 1Wednesday, 25 October 2023Work through the practical exercises "Getting Started" and "Step-by-step". Perform "Step-by-step" synthesis for at least 2 different sentences. Construct sentences that are short but contain plenty of ambiguity that needs resolving, so that as many of the steps as possible in Festival's front end actually do something interesting. Find at least one example where the ambiguity was correctly resolved, and one where it was not.
-
Speech Processing Lab - Group 1Wednesday, 25 October 2023 at 09:00 - 11:00Start TTS assignment
Before this lab: read Module 5 - speech synthesis – phonemes and the front end - Lab
Work through the practical exercises "Getting Started" and "Step-by-step" in the assignment specification https://speech.zone/exercises/the-festival-text-to-speech-system/.
Staff: Atli Sigurgeirsson, Simon King -
Speech Processing Lab - Group 2Wednesday, 25 October 2023 at 16:00 - 18:00Start TTS assignment
Before this lab: read Module 5 - speech synthesis – phonemes and the front end - Lab
Work through the practical exercises "Getting Started" and "Step-by-step" in the assignment specification https://speech.zone/exercises/the-festival-text-to-speech-system/.
Staff: Zihang Peng, Simon King
-
- Thursday, 26 October 2023
-
-
Lecture 6: Speech Synthesis - Waveform GenerationThursday, 26 October 2023 at 09:00 - 11:007GS_F.21, 7 George SquareManipulating recorded speech signals to create new utterances.
Before this class: complete Module 6
Lecturer: Catherine Lai
-
- Friday, 27 October 2023
-
-
Milestone - assignment 1Friday, 27 October 2023Find multiple errors in each of the categories required for the assignment.
-
- Monday, 30 October 2023
-
-
Assessment - TTS/Signals online quiz (10% of course mark)Due on Wednesday, 01 November 2023 by 12:00Multiple choice questions: Test open for 2 days, but once you start you need to complete within 1 hour on Learn. This test will cover material from modules 4-6.
-
- Tuesday, 31 October 2023
-
-
Assessment - TTS/Signals online quiz (10% of course mark)Due on Wednesday, 01 November 2023 by 12:00Multiple choice questions: Test open for 2 days, but once you start you need to complete within 1 hour on Learn. This test will cover material from modules 4-6.
-
- Wednesday, 01 November 2023
-
-
Milestone - assignment 1Wednesday, 01 November 2023Decide which 1-2 errors in each category (possibly more than 2 waveform generation errors) you will include in your report. Draft a good explanation of each of them. Assemble the evidence you need to present in the report, such as incorrect/correct pronunciations, annotated spectrograms, etc.
-
Speech Processing Lab - Group 1Wednesday, 01 November 2023 at 09:00 - 11:00Continue TTS assignment
Before this lab: read Module 6 - Speech Synthesis – waveform generation and connected speech - Lab, and bring a writing sample with you to the lab
Staff: Atli Sigurgeirsson, Simon King -
Assessment - TTS/Signals online quiz (10% of course mark)Due on Wednesday, 01 November 2023 by 12:00Multiple choice questions: Test open for 2 days, but once you start you need to complete within 1 hour on Learn. This test will cover material from modules 4-6.
-
Speech Processing Lab - Group 2Wednesday, 01 November 2023 at 16:00 - 18:00Continue TTS assignment
Before this lab: read Module 6 - Speech Synthesis – waveform generation and connected speech - Lab, and bring a writing sample with you to the lab
Staff: Zihang Peng, Simon King
-
- Tuesday, 07 November 2023
-
-
Assessment - submit assignment 1 (30% of course mark)Due on Tuesday, 07 November 2023 by 12:00via Learn
-
- Wednesday, 08 November 2023
-
-
Speech Processing Lab - Group 1Wednesday, 08 November 2023 at 09:00 - 10:50Command line and shell scripting
Before this lab: Work through the material in the Intermission module then come to the lab to get help.
Staff: Atli Sigurgeirsson, Simon King -
Speech Processing Lab - Group 2Wednesday, 08 November 2023 at 16:10 - 18:00Command line and shell scripting
Before this lab: Work through the material in the Intermission module then come to the lab to get help.
Staff: Zihang Peng, Simon King
-
- Thursday, 09 November 2023
-
-
Lecture 7: Speech Recognition - Pattern RecognitionThursday, 09 November 2023 at 09:00 - 10:507GS_F.21, 7 George SquareWe now start on an introduction to Automatic Speech Recognition, starting with the concept of pattern recognition.
Before this class: complete Module 7
Lecturer: Catherine Lai
-
- Friday, 10 November 2023
-
-
Milestone - assignment 2Friday, 10 November 2023Read through assignment 2 instructions and make sure you have a basic understanding of the linux command line by going through the LinkedIn online course from last week: https://www.linkedin.com/learning/learning-linux-command-line-2/?u%3D50251009%26auth%3Dtrue&sa=D&source=calendar&usd=2&usg=AOvVaw3E6yRZAbRovSbOoP8-Wj0c" target="_blank">https://www.linkedin.com/learning/learning-linux-command-line-2/?u=50251009&auth=true
-
- Wednesday, 15 November 2023
-
-
Milestone - assignment 2Wednesday, 15 November 2023Run the scripts given in the assignment to train a digit recogniser using data from a single user (e.g., simonk) by hand-coding that username into each script. Run the 4 scripts and obtain a WER for that user (for simonk you should get 0% WER). Try testing that recogniser on data from another user.
-
Speech Processing Lab - Group 1Wednesday, 15 November 2023 at 09:00 - 10:50Start ASR assignment Before this lab: read Module 7 - Speech Recognition – Pattern matching - Lab Staff: Atli Sigurgeirsson, Benjamin Elie, Simon King
-
Speech Processing Lab - Group 2Wednesday, 15 November 2023 at 16:10 - 18:00Start ASR assignment Before this lab: read Module 7 - Speech Recognition – Pattern matching - Lab Staff: Zihang Peng, Simon King
-
- Thursday, 16 November 2023
-
-
Lecture 8: Speech Recognition - Feature EngineeringThursday, 16 November 2023 at 09:00 - 10:507GS_F.21, 7 George SquareTo get the best out of machine learning, we can prepare features that reflect our knowledge of the problem, and suit our chosen model.
Before this class: complete Module 8
Lecturer: Catherine Lai
-
- Monday, 20 November 2023
-
-
Milestone - assignment 2Monday, 20 November 2023Improve the scripts so that they accept the user's name as a command line argument. Combine initialise_models and train_models into a single script called initialise_and_train_models (alternatively, initialise_and_train_models can run initialise_models then train_models). Modify recognise_test_data so that it first deletes any files left in the rec directory by a previous experiment.
-
- Wednesday, 22 November 2023
-
-
Milestone - assignment 2Wednesday, 22 November 2023Write a script called run_experiment that takes two command line arguments: the first is the name of a file containing the users in the training set, and the other is for the test set. The script should train, test and score a complete ASR system, by running initialise_and_train_models, then recognise_test_data, then results. It will print out the WER at the end. Use the forums to learn all the shell scripting techniques needed and keep asking questions whenever you get stuck. Design your first few experiments.
-
Speech Processing Lab - Group 1Wednesday, 22 November 2023 at 09:00 - 10:50Continue ASR assignment Before this lab: continue working in your own time on the assignment so that you have questions ready to ask in the lab Staff: Atli Sigurgeirsson, Simon King
-
Speech Processing Lab - Group 2Wednesday, 22 November 2023 at 16:10 - 18:00Continue ASR assignment
Before this lab: continue working in your own time on the assignment so that you have questions ready to ask in the lab
Staff: Zihang Peng, Simon King
-
- Thursday, 23 November 2023
-
-
Lecture 9: Speech Recognition - the Hidden Markov ModelThursday, 23 November 2023 at 09:00 - 10:507GS_F.21, 7 George SquareWe now replace pattern matching with a generative model that is learned from data.
Before this class: complete Module 9
Lecturer: Catherine Lai
-
- Friday, 24 November 2023
-
-
Milestone - assignment 2Friday, 24 November 2023Create all the training and test sets required for your experiments. Run all your isolated digits experiments. Start drafting your report, ready for the writing clinics next week.
-
- Monday, 27 November 2023
-
-
Milestone - assignment 2Monday, 27 November 2023Decide which aspect of your writing you need most help so you can ask for help from the PPLS skills centre or in the labs
-
Assessment - ASR online quiz (10% of course mark)Due on Wednesday, 29 November 2023 by 12:00Multiple choice questions: Test open for 2 days, but once you start you need to complete within 1 hour on Learn. This test will cover material from modules 7-9.
-
- Tuesday, 28 November 2023
-
-
Assessment - ASR online quiz (10% of course mark)Due on Wednesday, 29 November 2023 by 12:00Multiple choice questions: Test open for 2 days, but once you start you need to complete within 1 hour on Learn. This test will cover material from modules 7-9.
-
- Wednesday, 29 November 2023
-
-
Milestone - assignment 2Wednesday, 29 November 2023Finalise all the results tables and plots for your report, apart from digit sequences.
-
Speech Processing Lab - Group 1Wednesday, 29 November 2023 at 09:00 - 10:50Complete ASR assignment Before this lab: continue working in your own time on the assignment so that you have questions ready to ask in the lab Staff: Atli Sigurgeirsson, Benjamin Elie, Simon King
-
Assessment - ASR online quiz (10% of course mark)Due on Wednesday, 29 November 2023 by 12:00Multiple choice questions: Test open for 2 days, but once you start you need to complete within 1 hour on Learn. This test will cover material from modules 7-9.
-
Speech Processing Lab - Group 2Wednesday, 29 November 2023 at 16:10 - 18:00Complete ASR assignment
Before this lab: continue working in your own time on the assignment so that you have questions ready to ask in the lab
Staff: Zihang Peng, Simon King
-
- Thursday, 30 November 2023
-
-
Lecture 10: Speech Recognition - Connected speech & HMM trainingThursday, 30 November 2023 at 09:00 - 10:507GS_F.21, 7 George SquareHMMs extend easily to connected speech so finally we put everything together to make a complete speech recognition system. We'll also learn how to train an HMM from data.
Before this class: complete Module 10
Lecturer: Catherine Lai
-
- Friday, 01 December 2023
-
-
Milestone - assignment 2Friday, 01 December 2023If you've completed the rest of the assignment, create a language model for digit sequences. The techniques for this will have been taught in Module 10. Start building a system for digit sequences.
-
- Monday, 04 December 2023
-
-
Speech Processing Lab - Groups 1 & 2Monday, 04 December 2023 at 09:00 - 10:50Last chance for help with the ASR assignment
Both groups welcome (priority to Group 1 if the lab is full)
Staff: Atli Sigurgeirsson, Simon King -
Speech Processing Lab - Groups 2 & 1Monday, 04 December 2023 at 16:10 - 18:00Last chance for help with the ASR assignment Both groups welcome (priority to Group 2 if the lab is full) Staff: Zihang Peng, Simon King
-
- Wednesday, 06 December 2023
-
-
Milestone - assignment 2Wednesday, 06 December 2023Finish the system for digit sequences. Finalise your report, reading for final proof-reading and one last editing pass.
-
- Thursday, 07 December 2023
-
-
Assessment - submit assignment 2 (40% of course mark) - date subject to confirmationDue on Thursday, 07 December 2023 by 12:00via Learn
-