检索结果-内蒙古大学图书馆

Combining TF-GridNet And Mixture Encoder For Continuous Speech Separation For Meeting Transcription

学校读者我要写书评

暂无评论

Combining TF-GridNet And Mixture Encoder For Continuous Spee...

IEEE Spoken language technology Workshop

作者： Peter Vieting Simon Berger Thilo von Neumann Christoph Boeddeker Ralf Schlüter Reinhold Haeb-Umbach Machine Learning and Human Language Technology Group RWTH Aachen University Germany AppTek GmbH Germany Paderborn University Germany

ISBN: (数字)9798350392258

ISBN: (纸本)9798350392265

Many real-life applications of automatic speech recognition (ASR) require processing of overlapped speech. A common method involves first separating the speech into overlap-free streams on which ASR is performed. Recently, TF-GridNet has shown impressive performance in speech separation in real reverberant conditions. Furthermore, a mixture encoder was proposed that leverages the mixed speech to mitigate the effect of separation artifacts. In this work, we extended the mixture encoder from a static two-speaker scenario to a natural meeting context featuring an arbitrary number of speakers and varying degrees of overlap. We further demonstrate its limits by the integration with separators of varying strength including TF-GridNet. Our experiments result in a new state-of-the-art performance on LibriCSS using a single microphone. They show that TF-GridNet largely closes the gap between previous methods and oracle separation independent of mixture encoding. We further investigate the remaining potential for improvement.

关键词： Particle separators Conferences Encoding Data models Reverberation Speech processing Streams Microphones Automatic speech recognition

COMBINING TF-GRIDNET AND MIXTURE ENCODER FOR CONTINUOUS SPEECH SEPARATION FOR MEETING TRANSCRIPTION

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Vieting, Peter Berger, Simon von Neumann, Thilo Boeddeker, Christoph Schlüter, Ralf Haeb-Umbach, Reinhold Machine Learning and Human Language Technology Group RWTH Aachen University Germany AppTek GmbH Germany Paderborn University Germany

关键词： Continuous speech recognition

Comparative Analysis of the wav2vec 2.0 Feature Extractor

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Vieting, Peter Schlüter, Ralf Ney, Hermann Machine Learning and Human Language Technology RWTH Aachen University Germany AppTek GmbH Germany

Automatic speech recognition (ASR) systems typically use handcrafted feature extraction pipelines. To avoid their inherent information loss and to achieve more consistent modeling from speech to transcribed text, neural raw waveform feature extractors (FEs) are an appealing approach. Also the wav2vec 2.0 model, which has recently gained large popularity, uses a convolutional FE which operates directly on the speech waveform. However, it is not yet studied extensively in the literature. In this work, we study its capability to replace the standard feature extraction methods in a connectionist temporal classification (CTC) ASR model and compare it to an alternative neural FE. We show that both are competitive with traditional FEs on the LibriSpeech benchmark and analyze the effect of the individual components. Furthermore, we analyze the learned filters and show that the most important information for the ASR system is obtained by a set of bandpass filters. Copyright © 2023, The Authors. All rights reserved.

关键词： Speech recognition

An evolutionary confidence measurement for spoken term detection

学校读者我要写书评

暂无评论

An evolutionary confidence measurement for spoken term detec...

International Workshop on Content-Based Multimedia Indexing

作者： Tejedor, Javier Echeverría, Alejandro Wang, Dong Human Computer Technology Laboratory Universidad Autónoma de Madrid Spain Machine Learning Group Universidad Autónoma de Madrid Spain Multimedia Department Eurecom Sophia Antipolis France

ISBN: (纸本)9781612844336

We propose a new discriminative confidence measurement approach based on an evolution strategy for spoken term detection (STD). Our evolutionary algorithm, named evolutionary discriminant analysis (EDA), optimizes classification errors directly, which is a salient advantage compared with some conventional discriminative models which optimize objective functions based on certain class encoding, e.g. MLPs and SVMs. In addition, with the intrinsic randomness of the evolution strategy, EDA largely reduces the risk of converging to local minimums in model training. This is particularly valuable when the decision boundary is complex, which is the case when dealing with out-of-vocabulary (OOV) terms in STD. Experimental results on the meeting domain in English demonstrate considerable performance improvement with the EDA-based confidence for OOV terms compared with MLPs- and SVMs-based confidences;for in-vocabulary terms, however, no significant difference is observed with the three models. This confirms our conjecture that EDA exhibits more advantage for tasks with complex decision boundaries. © 2011 IEEE.

关键词： Discriminant analysis

Refined Statistical Bounds for Classification Error Mismatches with Constrained Bayes Error

学校读者我要写书评

暂无评论

Refined Statistical Bounds for Classification Error Mismatch...

IEEE (first ITW conference of the year) Information Theory Workshop (ITW)

作者： Zijian Yang Vahe Eminyan Ralf Schlüter Hermann Ney Computer Science Department Machine Learning and Human Language Technology Group Lehrstuhl Informatik 6 RWTH Aachen University Germany AppTek GmbH Germany

ISBN: (数字)9798350348934

ISBN: (纸本)9798350348941

In statistical classification/multiple hypothesis testing and machine learning, a model distribution estimated from the training data is usually applied to replace the unknown true distribution in the Bayes decision rule, which introduces a mismatch between the Bayes error and the model-based classification error. In this work, we derive the classification error bound to study the relationship between the Kullback-Leibler divergence and the classification error mismatch. We first reconsider the statistical bounds based on classification error mismatch derived in previous works, employing a different method of derivation. Then, motivated by the observation that the Bayes error is typically low in machine learning tasks like speech recognition and pattern recognition, we derive a refined Kullback-Leibler-divergence-based bound on the error mismatch with the constraint that the Bayes error is lower than a threshold.

关键词： Analytical models Conferences Training data machine learning Speech recognition Data models Bayes methods Testing Information theory

Dynamic Encoder Size Based on Data-Driven Layer-wise Pruning for Speech Recognition

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Xu, Jingjing Zhou, Wei Yang, Zijian Beck, Eugen Schlüter, Ralf Machine Learning and Human Language Technology Group Computer Science Dept. RWTH Aachen University Germany AppTek GmbH Aachen52062 Germany

Varying-size models are often required to deploy ASR systems under different hardware and/or application constraints such as memory and latency. To avoid redundant training and optimization efforts for individual models of different sizes, we present the dynamic encoder size approach, which jointly trains multiple performant models within one supernet from scratch. These subnets of various sizes are layer-wise pruned from the supernet, and thus, enjoy full parameter sharing. By combining score-based pruning with supernet training, we propose two novel methods, Simple-Top-k and Iterative-Zero-Out, to automatically select the best-performing subnets in a data-driven manner, avoiding resource-intensive search efforts. Our experiments using CTC on both Librispeech and TED-LIUM-v2 corpora show that our methods can achieve on-par performance as individually trained models of each size category. Also, our approach consistently brings small performance improvements for the full-size supernet. Copyright © 2024, The Authors. All rights reserved.

关键词：

Refined Statistical Bounds for Classification Error Mismatches with Constrained Bayes Error

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Yang, Zijian Eminyan, Vahe Schlüter, Ralf Ney, Hermann Machine Learning and Human Language Technology Group Lehrstuhl Informatik 6 Computer Science Department RWTH Aachen University Germany AppTek GmbH Germany

关键词： Adversarial machine learning

Automating airborne pollen classification: Identifying and interpreting hard samples for classifiers

学校读者我要写书评

暂无评论

Heliyon 2025年第2期11卷 e41656页

作者： Milling, Manuel Rampp, Simon D.N. Triantafyllopoulos, Andreas Plaza, Maria P. Brunner, Jens O. Traidl-Hoffmann, Claudia Schuller, Björn W. Damialis, Athanasios CHI – Chair of Health Informatics MRI Technical University of Munich Munich Germany MCML–Munich Center for Machine Learning Germany EIHW – Chair of Embedded Intelligence for Health Care & Wellbeing University of Augsburg Augsburg Germany Institute of Environmental Medicine and Integrative Health Faculty of Medicine University Clinic of Augsburg & University of Augsburg Augsburg Germany Institute of Environmental Medicine Helmholtz Center Munich German Research Center for Environmental Health Germany Faculty of Business and Economics and Faculty of Medicine University of Augsburg Augsburg Germany Department of Technology Management and Economics Technical University of Denmark Denmark Next Generation Technology Region Zealand Denmark Christine Kühne Center for Allergy Research and Education Davos Switzerland MDSI–Munich Data Science Institute Germany GLAM–the Group on Language Audio & Music Imperial College London London United Kingdom Terrestrial Ecology and Climate Change Department of Ecology School of Biology Faculty of Sciences Aristotle University of Thessaloniki Thessaloniki Greece

Deep-learning-based classification of pollen grains has been a major driver towards automatic monitoring of airborne pollen. Yet, despite an abundance of available datasets, little effort has been spent to investigate which aspects pose the biggest challenges to the (often black-box- resembling) pollen classification approaches. To shed some light on this issue, we conducted a sample-level difficulty analysis based on the likelihood for one of the largest automatically-generated datasets of pollen grains on microscopy images and investigated the reason for which certain airborne samples and specific pollen taxa pose particular problems to deep learning algorithms. It is here concluded that the main challenges lie in A) the (partly) co-occurring of multiple pollen grains in a single image, B) the occlusion of specific markers through the 2D capturing of microscopy images, and C) for some taxa, a general lack of salient, unique features. Our code is publicly available under https://***/millinma/SDPollen © 2025

关键词： Deep learning Pollen recognition Sample difficulty analysis

Evolutionary discriminative confidence estimation for spoken term detection

学校读者我要写书评

暂无评论

Evolutionary discriminative confidence estimation for spoken...

作者： Tejedor, Javier Echeverría, Alejandro Wang, Dong Vipperla, Ravichander Human Computer Technology Laboratory Universidad Autónoma de Madrid Madrid Spain Machine Learning Group Universidad Autónoma de Madrid Madrid Spain Multimedia Communications Department EURECOM Sophia-Antipolis France

Spoken term detection (STD) is the task of searching for occurrences of spoken terms in audio archives. It relies on robust confidence estimation to make a hit/false alarm (FA) decision. In order to optimize the decision in terms of the STD evaluation metric, the confidence has to be discriminative. Multi-layer perceptrons (MLPs) and support vector machines (SVMs) exhibit good performance in producing discriminative confidence;however they are severely limited by the continuous objective functions, and are therefore less capable of dealing with complex decision tasks. This leads to a substantial performance reduction when measuring detection of out-of-vocabulary (OOV) terms, where the high diversity in term properties usually leads to a complicated decision boundary. In this paper we present a new discriminative confidence estimation approach based on evolutionary discriminant analysis (EDA). Unlike MLPs and SVMs, EDA uses the classification error as its objective function, resulting in a model optimized towards the evaluation metric. In addition, EDA combines heterogeneous projection functions and classification strategies in decision making, leading to a highly flexible classifier that is capable of dealing with complex decision tasks. Finally, the evolutionary strategy of EDA reduces the risk of local minima. We tested the EDA-based confidence with a state-of-the-art phoneme-based STD system on an English meeting domain corpus, which employs a phoneme speech recognition system to produce lattices within which the phoneme sequences corresponding to the enquiry terms are searched. The test corpora comprise 11 h of speech data recorded with individual head-mounted microphones from 30 meetings carried out at several institutes including ICSI;NIST;ISL;LDC;the Virginia Polytechnic Institute and State University;and the University of Edinburgh. The experimental results demonstrate that EDA considerably outperforms MLPs and SVMs on both classification and confidence meas

关键词： Function evaluation