Browsing by Author "Erdem, Ernur Sonat"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • No Thumbnail Available
    Item
    Efficient Recognition of Human Emotional States from Audio Signals
    (2014) Erdem, Ernur Sonat; Sert, Mustafa; https://orcid.org/0000-0002-7056-4245; AAB-8673-2019
    Automatic recognition of human emotional states is an important task for efficient human-machine communication. Most of existing works focus on the recognition of emotional states using audio signals alone, visual signals alone, or both. Here we propose empirical methods for feature extraction and classifier optimization that consider the temporal aspects of audio signals and introduce our framework to efficiently recognize human emotional states from audio signals. The framework is based on the prediction of input audio clips that are described using representative low-level features. In the experiments, seven (7) discrete emotional states (anger, fear, boredom, disgust, happiness, sadness, and neutral) from EmoDB dataset, are recognized and tested based on nineteen (19) audio features (15 standalone, 4 joint) by using the Support Vector Machine (SVM) classifier. Extensive experiments have been conducted to demonstrate the effect of feature extraction and classifier optimization methods to the recognition accuracy of the emotional states. Our experiments show that, feature extraction and classifier optimization procedures lead to significant improvement of over 11% in emotion recognition. As a result, the overall recognition accuracy achieved for seven emotions in the EmoDB dataset is 83.33% compared to the baseline accuracy of 72.22%.
  • Thumbnail Image
    Item
    Ses sinyallerinde duygu tanıma ve geri erişimi
    (Başkent Üniversitesi Fen Bilimleri Enstitüsü, 2014) Erdem, Ernur Sonat; Sert, Mustafa
    Ses sinyalinde duygu tanıma özelikle, görsel bilginin kısıtlı ya da hiç olmadığı durumlarda önem kazanmaktadır. Bu çalışmada, tam ve genişletilebilir bir ses tabanlı duygu tanıma ve geri erişim çatısı önerilmiştir. Makine öğrenme yöntemi olarak Destek Vektör Makineleri (DVM) kullanılmış ve performansını artırmak amacıyla parametre optimizasyonu gerçekleştirilmiştir. Ses içerik analizlerinde, uygun pencere ve atlama sürelerine karar verebilmek için ampirik analizler gerçekleştirilmiştir. Çalışmada, gürbüz öznitelikler bulmak amacıyla, 20 ses özniteliği üzerinde, DVM kullanılarak kapsamlı analizler yapılmış ve sonuçlar değerlendirilmiştir. Ayrıca, ses sinyallerinin duygu-tabanlı geri erişimi için, nokta, aralık ve en yakın komşuluk olarak adlandırılan sorgu türleri geliştirilmiş ve geri erişim başarımları değerlendirilmiştir. Deneysel sonuçlara göre, sınıflandırıcı parametre optimizasyonu ve önerilen ses analiz yöntemleri, dayanak tanıma başarımlarını arttırmaktadır. Emotion recognition from audio signals become more of significance especially when visual information is limited or absent. In this study, a complete and extensible audio-based emotion recognition and retrieval framework is proposed. Support Vector Machine (SVM) is employed as the machine learning scheme and parameter optimization methods are carried out to improve the performance of the learner. In audio content analysis, empirical analyses are performed to decide the proper window and hop sizes. In the study, extensive analyses are conducted using 20 audio features with SVM classifier to determine robust audio features and to evaluate the results. In addition, flexible querying abilities, namely point, range, and nearest neighbor are developed and retrieval performance is evaluated for emotion-based retrieval of audio signals. Based on the experiments, parameter optimization of the classifier along with the proposed audio analysis methods improve the baseline recognition accuracy.