Text this: Speech Emotion Recognition Using a Multi-Time-Scale Approach to Feature Aggregation and an Ensemble of SVM Classifiers