Speech emotion recognition using semi-NMF feature optimization

Abstract

In recent times, much research is progressing forward in the field of speech emotion recognition (SER). Many SER systems have been developed by combining different speech features to improve their performances. As a result, the complexity of the classifier increases to train this huge feature set. Additionally, some of the features could be irrelevant in emotion detection and this leads to a decrease in the emotion recognition accuracy. To overcome this drawback, feature optimization can be performed on the feature sets to obtain the most desirable emotional feature set before classifying the features. In this paper, semi-nonnegative matrix factorization (semi-NMF) with singular value decomposition (SVD) initialization is used to optimize the speech features. The speech features considered in this work are mel-frequency cepstral coefficients, linear prediction cepstral coefficients, and Teager energy operator-autocorrelation (TEO-AutoCorr).This work uses k-nearest neighborhood and support vector machine (SVM) for the classification of emotions with a 5-fold cross-validation scheme. The datasets considered for the performance analysis are EMO-DB and IEMOCAP. The performance of the proposed SER system using semi-NMF is validated in terms of classification accuracy. The results emphasize that the accuracy of the proposed SER system is improved remarkably upon using the semi-NMF algorithm for optimizing the feature sets compared to the baseline SER system without optimization.

Harr-NMF Feature Extraction for Multilevel Educational Technology Teaching Big Data System

“…Specifically, most teaching resource management systems "emphasize construction but not management," lack management design, and ignore the sustainable development of teaching resources; the voluntary construction process is single, lacking the support of relevant norms and standards, and ignoring the real-time update and sustainability of resources; in terms of actual investment, the system construction cost is high, the degree of resource sharing is low, there is a lack of compatibility, and duplicate construction exists in large numbers. To address the above problems, this paper, on the one hand, stores massive heterogeneous teaching resources with unified standards, uses big data technology to manage massive teaching resources, improves the horizontal scalability of resources, and reduces the cost of the system in resource management and operation [21,22]. On the other hand, a well-designed retrieval mechanism for the application can parallelize the search in the massive teaching resources, such that users can quickly and accurately retrieve the resources they want and enhance the application effect of the system [23].…”

A Channel Selection Method for Emotion Recognition From EEG Based on Swarm-Intelligence Algorithms

“…In other words, each particle has a position which is a form of a binary vector X i . The positions and velocities formulated as equation (15,16);…”

“…Feature extraction, feature selection, and classification are the main stages, and the success of each step affects the performance of the entire system. The main issue is to find emotional salient features from several sources, analyzing feature sets [16], [17] to eliminate the irrelevant/unnecessary features and developing new classification frameworks to improve accuracies of existing classifiers [3], [18]. This study focuses on emotion recognition from EEG using band powers and phase-locking values as features and sophisticated feature selection method based on swarmintelligence (SI) algorithms and well-known classification algorithms such as k-nearest neighbour (k-NN), random forest, and support vector machines (SVM).…”

Classification of Traffic-Related Sounds Using Auditory Models and Convolutional Neural Networks

“…Bu sebeple, daha ekonomik ve müdahaleci olmayan trafik ölçümleri için Akustik Trafik İzleme (ATİ), son yıllarda popülerlik kazanmıştır. Gürültü kirliliği sorunu nedeniyle, son yıllarda ses ortamlarını izlemeye yönelik teknolojilerin geliştirilmesi hızlanmıştır [2] [3]. Otomatik ses tanıma (OST) teknolojisi, sinyal işleme ve makine öğrenimi tekniklerini kullanarak ses olaylarını otomatik olarak tanımlamak için önemli bir araç haline gelmiştir.…”

“…Bu yaklaşım, ses tanıma teknolojisinin geniş bir uygulama yelpazesine yayılmasına olanak tanır. Ses tanıma görevlerinde prozodik özellikler[1], ses kalitesi[2], Teager enerji operatörü (TEO)[3] ve spektral özellikler gibi birçok farklı özellik kullanılabilmektedir. Bununla birlikte son yıllarda geleneksel el ile seçilmiş özelliklerin yanı sıra 2 boyutlu zaman-frekans gösterimleri de yaygın olarak kullanılmaya başlanmıştır.…”