International audienceOne of the main limitations of functional connectivity estimators of brain networks is that they can suffer from statistical reliability when the number of areas is large and the available time series are short. To estimate directed functional connectivity with multivariate autoregressive (MVAR) model on sparse connectivity assumption, we propose a modified Group Lasso procedure with an adapted penalty. Our procedure includes the innovation estimates as explaining variables. This approach is inspired by two criteria that are used to interpret the coefficients of the MVAR model, the Directed Transfer Function (DTF) and the Partial Directed Coherence (PDC). A causality measure can be deduced from the output coefficients which can be understood as a synthesis of PDC and DTF. We demonstrate the potential of our method and compare our results with the standard Group Lasso on simulated data. © (2015) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only

In spite of its acoustic diversity, the speech signal presents statistical regularities that can be exploited by biological or artificial systems for efficient coding. Independent Component Analysis (ICA) revealed that on small time scales (* 10 ms), the overall structure of speech is well captured by a time-frequency representation whose frequency selectivity follows the same power law in the high frequency range 1-8 kHz as cochlear frequency selectivity in mammals. Variations in the power-law exponent, i.e. different time-frequency trade-offs, have been shown to provide additional adaptation to phonetic categories. Here, we adopt a parametric approach to investigate the variations of the exponent at a finer level of speech. The estimation procedure is based on a measure that reflects the sparsity of decompositions in a set of Gabor dictionaries whose atoms are Gaussian-modulated sinusoids. We examine the variations of the exponent associated with the best decomposition, first at the level of phonemes, then at an intra-phonemic level. We show that this analysis offers a rich interpretation of the fine-grained statistical structure of speech, and that the exponent values can be related to key acoustic properties. Two main results are: i) for plosives, the exponent is lowered by the release bursts, concealing higher values during the opening phases; ii) for vowels, the exponent is bound to formant bandwidths and decreases with the degree of acoustic radiation at the lips. This work further suggests that an efficient coding strategy is to reduce frequency selectivity with sound intensity level, congruent with the nonlinear behavior of cochlear filtering.

No abstract

Efficient coding of sensory signals takes advantage of statistical regularities in sensory data. Cochlear filter in mammals are known to reflect the overall statistical structure of speech, in line with the hypothesis that low-level sensory processing provides efficient codes for information contained in natural stimuli. Recently, some efforts have been made to describe this correspondence in more detail. The study of the statistical structure of speech over different acoustic classes demonstrates that frequency selectivity should not be fixed to achieve maximum efficiency. On the other hand, cochlear signal processing is nonlinear as frequency selectivity decreases with sound intensity level. Both effects are greater in the high frequencies. In the present study, these two facts are shown to be consistent in the case of a parametric method based on Gabor dictionaries (Gaussian-modulated sinusoids) and in a simplified setting. A model with fewer constrains is also introduced for future experiments to validate this hypothesis in a more general context.