Browsing by Author "Málek Jiří"
Now showing 1 - 20 of 25
Results Per Page
Sort Options
- Item2015 IEEE International Workshop of Electronics, Control, Measurement, Signals and their application to Mechatronics (ECMSM-2015)(2015-01-01) Plíva Zdeněk; Rozkovec Martin; Málek Jiří; Pfeifer Petr
- ItemAdaptive blind audio source extraction supervised by dominant speaker identification using x-vectors(IEEE, 2020-01-01) Janský Jakub; Málek Jiří; Čmejla Jaroslav; Kounovský Tomáš; Koldovský Zbyněk; Žďánský Jindřich
- ItemAdvanced semi-blind speaker extraction and tracking implemented in experimental device with revolving dense microphone array(International Speech Communication Association, 2021-01-01) Čmejla Jaroslav; Kounovský Tomáš; Janský Jakub; Málek Jiří; Rozkovec Martin; Koldovský Zbyněk
- ItemAuxiliary Function-Based Algorithm for Blind Extraction of a Moving Speaker(Springer, 2021-01-01) Janský Jakub; Koldovský Zbyněk; Málek Jiří; Kounovský Tomáš; Čmejla Jaroslav
- ItemBlind extraction of moving audio source in a challenging environment supported by speaker identification via X-vectors(IEEE, 2021-01-01) Málek Jiří; Janský Jakub; Kounovský Tomáš; Koldovský Zbyněk; Žďánský Jindřich
- ItemBlock-Online Multi-Channel Speech Enhancement Using DNN-Supported Relative Transfer Function Estimates(Institution of Engineering and Technology, 2020-01-01) Málek Jiří; Koldovský Zbyněk; Boháč Marek
- ItemCHiME4: Multichannel Enhancement Using Beamforming Driven by a DNN-based Voice Activity Detector(2016-01-01) Koldovský Zbyněk; Málek Jiří; Boháč Marek; Janský Jakub
- ItemCompensation of Nonlinear Distortions in Speech for Automatic Recognition(Institute of Electrical and Electronics Engineers Inc., 2015-01-01) Málek Jiří; Silovský Jan; Červa Petr; Koldovský Zbyněk; Nouza Jan; Žďánský Jindřich
- ItemDynamic Independent Component/Vector Analysis: Time-Variant Linear Mixtures Separable by Time-Invariant Beamformers(IEEE, 2021-01-01) Koldovský Zbyněk; Kautský Václav; Tichavský Petr; Čmejla Jaroslav; Málek Jiří
- ItemExtraction of Independent Vector Component from Underdetermined Mixtures through Block-Wise Determined Modeling(IEEE, 2019-01-01) Koldovský Zbyněk; Málek Jiří; Janský Jakub
- ItemHammerstein Model-Based Nonlinear Echo Cancelation Using a Cascade of Neural Network and Adaptive Linear Filter(Institute of Electrical and Electronics Engineers Inc., 2016-01-01) Málek Jiří; Koldovský Zbyněk
- ItemImproving relative transfer function estimates using second-order cone programming(Springer Verlag, 2015-01-01) Koldovský Zbyněk; Málek Jiří; Tichavský Petr
- ItemIndependent Vector Analysis Exploiting Pre-Learned Banks of Relative Transfer Functions for Assumed Target8217s Positions(2018-01-01) Čmejla Jaroslav; Kounovský Tomáš; Málek Jiří; Koldovský Zbyněk
- ItemNonlinear echo cancellation using generalized power filters(Institute of Electrical and Electronics Engineers Inc., 2015-01-01) Málek Jiří; Koldovský Zbyněk
- ItemPhone Speech Detection and Recognition in the Task of Historical Radio Broadcast Transcription(Institute of Electrical and Electronics Engineers Inc., 2015-01-01) Chaloupka Josef; Nouza Jan; Málek Jiří; Silovský Jan
- ItemRobust Automatic Recognition of Speech with Background Music(Institute of Electrical and Electronics Engineers Inc., 2017-01-01) Málek Jiří; Žďánský Jindřich; Červa PetrThis paper addresses the task of Automatic Speech Recognition (ASR) with music in the background, where the accuracy of recognition may deteriorate significantly. To improve the robustness of ASR in this task, e.g. for broadcast news transcription or subtitles creation, we adopt two approaches: 1) multi-condition training of the acoustic models and 2) denoising autoencoders followed by acoustic model training on the preprocessed data. In the latter case, two types of autoencoders are considered: the fully connected and the convolutional network. Presented experimental results show that all the investigated techniques are able to improve the recognition of speech distorted by music significantly. For example, in the case of artificial mixtures of speech and electronic music (low Signal-to-Noise Ratio (SNR) of 0 dB), we achieved absolute improvement of accuracy by 35.8%. For real-world broadcast news and a high SNR (about 10 dB), we achieved improvement by 2.4%. The important advantage of the studied approaches is that they do not deteriorate the accuracy in scenarios with clean speech (the decrease is about 1%).
- ItemRobust Recognition of Conversational Telephone Speech via Multi-Condition Training and Data Augmentation(Springer Verlag, 2018-01-01) Málek Jiří; Žďánský Jindřich; Červa Petr
- ItemRobust Recognition of Speech with Background Music in Acoustically Under-Resourced Scenarios(IEEE, 2018-01-01) Málek Jiří; Žďánský Jindřich; Červa Petr
- ItemSingle channel speech enhancement using convolutional neural network(2017-01-01) Kounovský Tomáš; Málek Jiří
- ItemSpatial source subtraction based on incomplete measurements of relative transfer function(Institute of Electrical and Electronics Engineers Inc., 2015-01-01) Koldovský Zbyněk; Málek Jiří; Gannot S.