Roman Jarina

A QoE Evaluation Methodology for HD Video Streaming Using Social Networking

A novel methodology for QoE evaluation in the social network environment is proposed. It provides high applicability for subjective testing of the multimedia services with respect to real usage scenarios. The environment of social networks provides also significant demographic data and ability to contact extremely many test subjects while allows to focus on or filter specific social groups. QoE results for HD internet video services are presented and followed by discussion on their statistical significance.

Model parameters selection for SVM classification using Particle Swarm Optimization

Support Vector Machine (SVM) classification requires set of one or more parameters and these parameters have significant influence on classification precision and generalization ability. Searching for suitable model parameters invokes great computational load, which accentuates with increasing size of the dataset and with amount of the parameters being optimized. In this paper we present and compare various SVM parameters selection techniques, namely grid search, Particle Swarm Optimization (PSO) and Genetic Algorithm (GA). Experiments conducting over two datasets show promising results with PSO and GA optimization technique.

Development of a Reference Platform for Generic Audio Classification

Detection of key sounds, such as applause, laugh, music, environmental noise, etc., is one of the challenges in intelligent management of multimedia information and content understanding. In this paper, we report progress in development of a reference content-based audio classification algorithm that is based on a conventional and widely accepted approach, namely signal parameterization by MFCC followed by GMM classification. Our developed labeled audio database and the conventional classification model should serve as a reference platform for an evaluation of novel, alternative or more advanced methods in audio content analysis.

GA-Based Feature Extraction for Clapping Sound Detection

Automatically extracting semantic content from audio streams can be helpful in many multimedia applications. In this paper, we introduce a framework for automatic feature subspace selection from a common feature vector. The selected features build a new representation which is better suitable for a given learning task and recognition. In order to solve this problem, we propose the GA-based (genetic algorithm) method to improve the representativeness and robustness of the features generic audio recognition task

Compact representation of speech using 2-d cepstrum – an application to slovak digits recognition

HMM speech recogniser with a small number of acoustic observations based on 2-D cepstrum (TDC) is proposed. TDC represents both static and dynamic features of speech implicitly in matrix form. It is shown that TDC analysis enables a compact representation of speech signals. Thus a great advantage of the proposed model is a massive reduction of speech features used for recognition what lessens computational and memory requirements, so it may be favourable for limited-power ASR applications. Experiments on isolated Slovak digits recognition task show that the method gives comparable results as the conventional MFCC approach. For speech degraded by additive white noise, it reaches better performance than the MFCC method.

Assessment of automatic speaker verification on lossy transcoded speech

In this paper, we investigate the effect of lossy speech compression on text-independent speaker verification task. We have evaluated the voice biometrics performance over several state-of-the art speech codecs including recently released Enhanced Voice Services (EVS) codec. The tests were performed in both codec-matched and codec-mismatched scenarios. The test results show that EVS outperforms other speech codecs used in our test and it can be used to generate speaker models that are quite robust to varying compression levels. It was also shown that if a speech codec of higher quality (EVS, G711) is included in training data (mismatched and partially mismatched scenarios), the automatic speaker verification (ASV) gives better results than in the case of matched scenario.

SVM based speaker emotion recognition in continuous scale

In this paper we propose a system of speaker emotion recognition based on the SVM regression. Recognized emotional state is expressed in continuous scale in three dimensions: valence, activation and dominance. Experiments have been performed on the IEMOCAP database that contains 6 basic emotions supplemented with 3 additional emotions. Audio recordings from the corpus were divided into voiced and unvoiced segments, and for both types, a vast collection of diverse audio features (830/710) were extracted. Then 40 features for each type of segment were selected by Particle Swarm Optimization. Classification accuracy is expressed by cross-correlation coefficients between the estimated (by the propose system) and real (assigned according to human judgements) emotional state labels. Experiments conducted over dataset show very promising results for the future experiments.

Depth map computation using hybrid segmentation algorithm

In this paper, a stereo matching algorithm based on image segments is presented. We propose the hybrid segmentation algorithm that is based on a combination of the Belief Propagation and Mean Shift algorithms with aim to refine the depth map by using a stereo pair of images. Firstly, a color based segmentation method is applied for segmenting the left image of the input stereo pair (reference image) into regions. The aim of the segmentation is to simplify representation of the image into the form that is easier to analyze and is able to locate objects in images. Secondly, results of the segmentation are used as an input of the local window-based matching method to determine the disparity estimate of each image pixel. The obtained experimental results demonstrate that the final depth map can be obtained by application of segment disparities to the original images.

Influence of packet loss on a speaker verification system over IP network

The paper considers an influence of packet loss on a remote speaker verification in Voice over IP (VoIP) environment. A lossy speech coding and packet loss represent a significant part of speech degradation in the VoIP environment. As an extent of packet loss impact is tightly related to a type of speech coder used to transmit speech data, different transmission conditions along with different speech codecs are investigated here. The speaker verification system used in this experimental study is based on a probabilistic GMM-UBM approach. In this paper, a speaker verification accuracy is evaluated against a level of packet loss in narrowband and wideband communication channel.

A novel approach for 3D model recognition based on SSCD

In this paper, a 3D model recognition method based on modification of Spatial Structure Circular Descriptor (SSCD) is proposed. Firstly, the model optimization process removes all valueless points. Secondly, the SSCD descriptor to get a spatial value is used. This method uses a Spherical Grade Projection (SGP) to project points on a plane. However, in our case the SGP is not used. Our proposed method is based on extraction of spatial information, which is projected to image. To store the full spatial information of 3D model to an image spherical transformation is used. Mostly, low-polygon models have small amount of points, which are projected to sphere hence, is created empty spaces. To avoid this problem is used a convolution with gradient function. Finally, we calculate the similarities between dataset of 3D models. The algorithm has been tested on 100 different 3D models (10 models for each class). The experimental result shows that the proposed method has a positive effect on overall recognition performance and outperforms other examined methods.