Cambridge Speech Decoding Workshop

Neuroimaging studies of auditory and speech perception have advanced from tightly controlled experiments in which short stimuli (isolated sounds, single words, etc.) are presented hundreds of times into more naturalistic studies with continuous tasks, such as listening to podcasts or spoken stories. This shift has been paralleled and facilitated by improvements in computational resources, signal processing, and machine learning methods, enabling new analysis methods for quantifying continuous brain activity in naturalistic communicative conditions. Such methods include temporal response functions, mutual information, and cross-correlation, for instance. Each share the underlying objective of quantifying systematic correspondences between continuous stimulus presentation and concurrent, continuously recorded brain activity measures. 

Although neural decoding is increasingly popular, it can be challenging to know which approaches or parameters are appropriate to the research questions in focus. Moreover, the diversity, complexity, and computational demands of these techniques is steadily increasing. This presents a hurdle for researchers seeking to gain a comprehensive understanding of these methods and to gain experience in their implementation. 

In response to this need, this workshop aims to provide attendees with an overview of the latest developments in both the theoretical and practical aspects of these modelling techniques. The goal is to help participants get started with and gain confidence in the application of these advanced methods—whether as a student questioning which technique is appropriate for their research question, or as a more experienced researcher seeking clarification concerning the differences between analysis methods, or choices made during data collection and preprocessing. 

10:00AM Welcome and coffee 

10:30AM Introduction - Why neural speech decoding?
Matt Davis (MRC-CBU, Cambridge, UK)

11:00AM Session 1 - Temporal Response Function methods
Marlies Gilllis (Experimental Otorhinolaryngology, KU Leuven, BE)

'Neural Speech Tracking: Selecting Models and Features'
Nathaniel Zuk (Psychology, Nottingham Trent University, UK)

'Interpreting the TRF: Simulations and considerations'

12:30PM Lunch break and Poster session

2:30PM Session 2 - Mutual Information methods
Hyojin Park (Psychology, University of Birmingham, UK)

'Introduction to Mutual Information & its Applications in Speech Tracking'

Christoph Daube (Psychology, University of Glasgow, UK) 

'A helicopter flies over a stadium and tries to figure out  what's going on'

4:00PM Discussion and next steps

4:30PM Finish. Drinks in the CBU garden if weather permits

Pub trip to The Grain & Hop Store 69 Regent St, Cambridge 

This workshop is funded by the Cambridge Language Sciences Interdisciplinary Research Centre.  

Attendance will be free for students and staff of the University of Cambridge (please  contact the organisers to get the password to be able to register for free using the "Uni of Cambridge Students & ECR without funding" option ). 

However, we ask that attendees with suitable grant funding contribute to lunch costs by selecting the "Uni of Cambridge Student/Staff: £20 Internal cross-charge" option. 

For non-Cambridge attendees there will be a limited number of places available on a first-come first-served basis in return for a small registration fee to cover catering costs (please select the "Standard Registration" option). 

If you wish to present a poster, discuss data or share work-in-progress during the lunchtime session, please contact the organisers. 

Register