Speech Emotion Recognition Based on Selective Interpolation Synthetic Minority Over-Sampling Technique in Small Sample Environment.

Speech emotion recognition often encounters the problems of data imbalance and redundant features in different application scenarios. Researchers usually design different recognition models for different sample conditions. In this study, a speech emotion recognition model for a small sample environment is proposed. A data imbalance processing method based on selective interpolation synthetic minority over-sampling technique (SISMOTE) is proposed to reduce the impact of sample imbalance on emotion recognition results. In addition, feature selection method based on variance analysis and gradient boosting decision tree (GBDT) is introduced, which can exclude the redundant features that possess poor emotional representation. Results of experiments of speech emotion recognition on three databases (i.e., CASIA, Emo-DB, SAVEE) show that our method obtains average recognition accuracy of 90.28% (CASIA), 75.00% (SAVEE) and 85.82% (Emo-DB) for speaker-dependent speech emotion recognition which is superior to some state-of-the-arts works.

Dynamic Network Connectivity Analysis to Identify Epileptogenic Zones Based on Stereo-Electroencephalography.

Objectives: Accurate localization of epileptogenic zones (EZs) is essential for successful surgical treatment of refractory focal epilepsy. The aim of the present study is to investigate whether a dynamic network connectivity analysis based on stereo-electroencephalography (SEEG) signals is effective in localizing EZs. Methods: SEEG data were recorded from seven patients who underwent presurgical evaluation for the treatment of refractory focal epilepsy and for whom the subsequent resective surgery gave a good outcome. A time-variant multivariate autoregressive model was constructed using a Kalman filter, and the time-variant partial directed coherence was computed. This was then used to construct a dynamic directed network model of the epileptic brain. Three graph measures (in-degree, out-degree, and betweenness centrality) were used to analyze the characteristics of the dynamic network and to find the important nodes in it. Results: In all seven patients, the indicative EZs localized by the in-degree and the betweenness centrality were highly consistent with the clinically diagnosed EZs. However, the out-degree did not indicate any significant differences between nodes in the network. Conclusions: In this work, a method based on ictal SEEG signals and effective connectivity analysis localized EZs accurately. The results suggest that the in-degree and betweenness centrality may be better network characteristics to localize EZs than the out-degree.

Localization of epileptogenic zone based on graph analysis of stereo-EEG.

Localization of the epileptogenic zone (EZ) is essential for the successful surgical treatment of medically intractable epilepsy. In the present study, stereo-EEG (SEEG) recordings were obtained from seven patients underwent presurgical evaluation for treatment of intractable epilepsy. Partial directed coherence (PDC) analysis was applied to construct peri-ictal effective connectivity networks. The graphic measures, in-degree, out-degree and betweenness centrality, were evaluated to localize the EZ. A receiver operating characteristic (ROC) analysis was used to quantify the localization accuracy. We found that the in-degree coincided well with the EZ identified by epileptologists' visual inspection in all seven patients who had a significant improvement in seizure outcomes, however, the other two measures were effective only in some cases. Furthermore, in all seven patients the electrode contact with the highest in-degree was always located within the EZ identified by epileptologists' visual inspection. These results indicate that the graph theory is an effective method to localize the EZ when suitable graphic measures were chosen. Furthermore, the in-degree was the most effective measure among the three graphic measures in localizing the EZ when the PDC method was used.