Speech processing

Speech processing, with its temporal cadence and multi-scale syntactic invariants (syllables, words), is a paradigmatic example where rhythms have been proposed to play a key role [1], with a speech-modulated hierarchical structure of intercoupled cortical oscillations correlating with successful comprehension. Experiments show that speech recognition remains largely intact when compressed up to a certain temporal factor [2] and the re-spacing of chunks of incomprehensible compressed speech with silences recovers comprehension. We hypothesize that rhythm-based top-down semantic context inferen

The human brain processes speech and melody differently, with the left hemisphere predominantly processing temporal features and the right hemisphere processing spectral features. This asymmetric processing raises questions about the underlying neural dynamics of acoustic feature encoding for speech and music. To further investigate this, we recorded intracranial EEG data from seventeen epileptic patients with implants in primary and associative auditory cortical regions. We utilized a stimulus set comprising one hundred a cappella songs, each presented in three versions: temporally degraded,

…Speech processing, with its temporal cadence and multi-scale of syntactic invariants (syllables, words), is a paradigmatic example where rhythms have been proposed to play a key role [1], with a speech-modulated hierarchical structure of intercoupled cortical oscillations correlating with successful comprehens

Understanding speech in noisy backgrounds requires selective attention to a particular speaker. Humans excel at this challenging task, while current speech recognition technology still struggles when background noise is loud. The neural mechanisms by which we process speech remain, however, poorly understood, not least due to the complexity of natural speech. Here we describe recent progress obtained through applying machine-learning to neuroimaging data of humans listening to speech in different types of background noise. In particular, we develop statistical models to relate characteristic f

Irina is a research scientist at DeepMind, where she works in the Froniers team. Her work aims to bring together insights from the fields of neuroscience and physics to advance general artificial intelligence through improved representation learning. Before joining DeepMind, Irina was a British Psychological Society Undergraduate Award winner for her achievements as an undergraduate student in Experimental Psychology at Westminster University, followed by a DPhil at the Oxford Centre for Computational Neuroscience and Artificial Intelligence, where she focused on understanding the computationa