Yazar "Arslan, Recep Sinan" seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • Küçük Resim
    Öğe
    Multi-lingual Speech Emotion Recognition System Using Machine Learning
    (Selçuk Üniversitesi, 2024) Çolakoğlu, Emel; Hızlısoy, Serhat; Arslan, Recep Sinan
    Predicting emotions from speech in different languages with high accuracy has been a challengingtask for researchers in recent years. When we delve into the studies conducted in this field, it isclear that researchers generally try to recognize emotions from speech in their traditional language.However, these studies cannot be generalized for multi-lingual environments around the globe.The Turkish speech emotional dataset, which was created for use in our previous studies, wasfurther expanded for use in this study too. Emo-db dataset was also used to benchmark the successof the proposed model. Various pre-processing stages such as standardization, sorting andresampling were applied to the data in the datasets to increase the performance of the model.OpenSMILE toolbox, which is frequently encountered in studies, was used to obtain features thatprovide meaningful information corresponding to the emotion in speech, and thousands of featureswere obtained from emobase2010 and emo_large feature sets. 8 different machine learningalgorithms were used in the model to classify 4 different emotions for the Turkish dataset and 7different emotions for the Emo-db dataset. The best recognition rates were achieved with 92.73%and 96.3%, respectively, for the Turkish dataset consisting of 1099 records and the Emo-db datasetconsisting of 535 records, using the Emobase2010 as a feature set and Logistic Regression as aclassifier.
  • Küçük Resim
    Öğe
    Text independent speaker recognition based on MFCC and machine learning
    (Selçuk Üniversitesi, 2021) Hizlisoy, Serhat; Arslan, Recep Sinan
    Speaker recognition (SR) is the process of recognizing the voice of human from a group of speech samples with artificial intelligence. SR models are used in various human-voice based security platforms and authentication problems. In this paper, a text-independent speaker recognition model was developed for the problem with 60 different speakers. Obtaining the distinctive features of speaker expressions during the model design phase is an important point. In this study, the MFCC algorithm, which is the most common method used to obtain short-time features, is used to extract features of speech signals. The classification performance of the proposed model and commonly used 11 different machine learning methods has been evaluated on Audio-MNIST dataset, and the results were shown comparatively. As a result, 97.1% classification rate was achieved with SVM classifier. In addition, precision, recall and f-score values are 98.0%, 97.1% and 97.4%, respectively. The results show that the proposed model produces successful results for all classes and is a widely applicable approach to different types of speaker datasets.