Analysis of singing during epileptic seizures based on partial coherence

This paper deals with the analysis of singing during epileptic seizures. This phenomenon is rare and seems to appear when the origin of the seizure is in the temporal lobe. Analyzing the relations between signals belonging to different structures may contribute to define the role of each of them and derive a neural network involved during singing production. To study the relations along time, we choose to estimate the coherence function and the partial coherence function. The results obtained show that some coherences increase during singing phenomenon, and differences appear between given coherences and the same coherences conditioned on a third signal. This last result informs us of the contribution of a signal in the relation between two others.