This paper deals with the problem of enhancing speech signals that have been degraded by statistically independent quasistationary noise. The estimation of the clean speech waveform, and of the parameters of autoregressive (AR) models for the clean speech, given the noisy speech, is considered. The two problems are demonstrated to be closely related in the sense that a good solution to one of them can be used for achieving a satisfactory solution for the other. The difficulties in solving these estimation problems are mainly due to the lack of explicit knowledge of the statistics of the clean speech signal and of the noise process. Maximum likelihood estimation solutions that are based upon the E–M algorithm and its derivatives are proposed. For estimating the speech waveform, the statistics of the clean speech signal and of the noise process are first estimated by training a pair of Gaussian AR hidden Markov models, one for the clean speech and the other for the noise, using long training sequences from the two sources. Then, the speech waveform is reestimated by applying the E–M algorithm to the estimated statistics. An approximation to the E–M algorithm is interpreted as being an iterative procedure in which Wiener filtering and AR modeling are alternatively applied. The different algorithms considered here will be compared and demonstrated.

Statistical model-based speech enhancement systems

Monaural multi-talker speech recognition using factorial speech processing models

De entre los problemas que aun se encuentran abiertos en el campo del reconocimiento automatico del habla, uno de los que suscitan mayor interes entre la comunidad cientifica es el robustecimiento de estos sistemas frente al ruido acustico. Es bien conocido que, en presencia de ruido, el rendimiento de estos sistemas se deteriora hasta niveles en los que su uso resulta ineficaz. A fin de mitigar este problema, en este trabajo se desarrollan un conjunto de tecnicas que permiten incrementar la robustez de estos sistemas en condiciones ruidosas. Para alcanzar esta meta, en esta tesis se adopta un marco de trabajo en el que las caracteristicas de voz usadas por el reconocedor son procesadas para mitigar la degradacion producida por el ruido. Bajo este marco generico, a lo largo de este trabajo proponemos varias tecnicas que, usando estimacion bayesiana MMSE e informacion a priori sobre la voz y/o el ruido, permiten obtener unas caracteristicas mas limpias. En primer lugar, se proponen un conjunto de tecnicas que, a partir de datos estereo, derivan una serie de transformaciones que se aplican a la voz ruidosa para compensarla. Estos datos estereo consisten en grabaciones de voz donde se cuenta con senales de voz limpia y sus correspondientes versiones ruidosas. Usando estas grabaciones, las tecnicas propuestas estiman un conjunto de parametros que se emplean posteriormente para realzar las caracteristicas extraidas de la voz ruidosa. A fin de eludir la necesidad de datos estereo de las tecnicas anteriores, en segundo lugar se proponen un conjunto de tecnicas de reconstruccion basadas en un modelo de enmascaramiento de las caracteristicas espectrales de voz. De acuerdo a este modelo, el ruido acustico enmascara (distorsiona) ciertas caracteristicas del espectro de la voz dejando otras intactas. Para recuperarse de esta perdida de informacion, se proponen dos tecnicas alternativas. La primera de ellas, denominada TGI (Truncated-Gaussian based Imputation), estima el valor de las regiones enmascaradas del espectro supuesto que se conoce por adelantado la clasificacion entre regiones limpias y ruidosas de la senal observada. La segunda tecnica, conocida como MMSR (Masking-Model based Spectral Reconstruction), emplea modelos a priori de voz y ruido para llevar a cabo esta tarea. Como se vera, la reconstruccion obtenida por MMSR se reduce a una combinacion lineal entre el espectro original y un espectro estimado en el que se consideran los posibles efectos del enmascaramiento del ruido. Basandonos en la formulacion desarrollada para la tecnica MMSR, tambien se propondra un algoritmo iterativo para la estimacion del modelo de ruido empleado por esta tecnica. Este algoritmo permite ajustar los parametros de un modelo de mezcla de gaussianas al ruido acustico presente en cada elocucion que se reconoce, consiguiendo de esta forma modelar las caracteristicas variantes de este tipo de ruidos. Asimismo, de esta formulacion tambien se derivara un algoritmo para la estimacion de las mascaras empleadas para identificar las regiones limpias y ruidosas del espectro de voz. Estas mascaras pueden emplearse por TGI y otras tecnicas similares. Finalmente, en esta tesis tambien se exploran otros dos cuestiones importantes para la compensacion de caracteristicas: el modelado temporal de la voz y el tratamiento de la incertidumbre. En relacion a la primera cuestion, se proponen dos alternativas para representar la evolucion temporal de la voz en los modelos usados por las tecnicas de compensacion propuestas: modelado de segmentos de voz y modelado basado en modelos ocultos de Markov. En cuanto a la segunda cuestion, se investiga el computo de medidas que describan la incertidumbre residual del proceso de compensacion de caracteristicas de voz, asi como su posterior uso en el reconocedor.

Reconocimiento robusto de voz con datos perdidos o inciertos

Model compensation methods for noise-robust speech recognition have shown good performance. Predictive linear transformations can approximate these methods to balance computational complexity and compensation accuracy. This paper examines both of these approaches from a variational perspective. Using a matched-pair approximation at the component level yields a number of standard forms of model compensation and predictive linear transformations. However, a tighter bound can be obtained by using variational approximations at the state level. Both model-based and predictive linear transform schemes can be implemented in this framework. Preliminary results show that the tighter bound obtained from the state-level variational approach can yield improved performance over standard schemes.

/pdf/a-variational-perspective-on-noise-robust-speech-recognition-bftsmp6osw.pdf

A variational perspective on noise-robust speech recognition

This paper investigates the effectiveness of factorial speech processing models in noise-robust automatic speech recognition tasks. For this purpose, the paper proposes an idealistic approach for modeling state-conditional observation distribution of factorial models based on weighted stereo samples. This approach is an extension to previous single-pass retraining for ideal model compensation which is extended here to support multiple audio sources. Non-stationary noises can be considered as one of these audio sources with multiple states. Experiments of this paper over the set A of the Aurora 2 dataset show that recognition performance can be improved by this consideration. The improvement is significant in low signal-to-noise energy conditions, up to 4 % absolute word recognition accuracy. In addition to the power of the proposed method in accurate representation of state-conditional observation distribution, it has an important advantage over previous methods by providing the opportunity to independently select feature spaces for both source and corrupted features. This opens a new window for seeking better feature spaces appropriate for noisy speech, independent from clean speech features.

/pdf/erratum-to-modeling-state-conditional-observation-4x3e5hx6d3.pdf

Erratum to: Modeling State-Conditional Observation Distribution Using Weighted Stereo Samples for Factorial Speech Processing Models

A standard way of improving the robustness of speech recognition systems to noise is model compensation. is replaces a speech recogniser’s distributions over clean speech by ones over noise-corrupted speech. For each clean speech component,model compensation techniques usually approximate the corrupted speech distribution with a diagonal-covariance Gaussian distribution. is thesis looks into improving on this approximation in two ways: rstly, by estimating full-covariance Gaussian distributions; secondly, by approximating corrupted-speech likelihoods without any parameterised distribution. e rst part of this work is about compensating for within-component feature correlations under noise. For this, the covariancematrices of the computed Gaussians should be full instead of diagonal. e estimation of o-diagonal covariance elements turns out to be sensitive to approximations. A popular approximation is the one that state-of-the-art compensation schemes, likevts compensation, use for dynamic coecients: the continuous-time approximation. Standard speech recognisers contain both per-time slice, static, coecients, and dynamic coecients, which represent signal changes over time, and are normally computed from a window of static coecients. To remove the need for the continuous-time approximation, this thesis introduces a new technique. It rst compensates a distribution over the window of statics, and then applies the same linear projection that extracts dynamic coecients. It introduces a number ofmethods that address the correlation changes that occur in noisewithin this framework. e next problem is decoding speed with full covariances. is thesis reanalyses the previously-introduced predictive linear transformations, and shows how they can model feature correlations at low and tunable computational cost. e second part of this work removes the Gaussian assumption completely. It introduces a sampling method that, given speech and noise distributions and a mismatch function, in the limit calculates the corrupted speech likelihood exactly. For this, it transforms the integral in the likelihood expression, and then applies sequential importance resampling. ough it is too slow to use for recognition, it enables a more ne-grained assessment of compensation techniques, based on the kl divergence to the ideal compensation for one component. e kl divergence proves to predict the word error rate well. is technique also makes it possible to evaluate the impact of approximations that standard compensation schemes make.

/pdf/statistical-models-for-noise-robust-speech-recognition-4flx20hsbl.pdf

Statistical models for noise-robust speech recognition

Handle everyday research tasks with reliable, citation-backed results

Your personal Research Agent to handle research tasks with citation-backed results

Popular Tasks used by Researchers

How can I help with your research?

Meet SciSpace

Get more enhanced response by uploading the PDFs you want me to reference.

No relevant PDFs in your library

SciSpace is the AI research assistant for academics. Run systematic literature reviews on 280M+ papers, and write papers with cited sources. Trusted by 1M+ students, PhDs & researchers.

SciSpace | AI for Research

Analyze PDFs

Code & Manuscripts

Funding & Grants

Literature & Patents

Medical & Clinical Data

Systematic Review

Visualize & Present

Web & Data

Build a Google Scholar-like website for your research.

Build a website

Create charts and images for your research

Create a Chart

Write a paper for submission to a journal

Draft a manuscript

Patent Search

Design eye-catching scientific posters in minutes.

Scientific Poster Generation

Systematic Literature Review

One task is running at the moment. Your messages will be shown right after.

Drag and drop or click here to browse

Loved by <highlight>1 million+</highlight> researchers

Extract a list of specific topics and their sources from unstructured text

Topics

Compare and analyze relevant papers that matches with your search

Papers

Get insights from PDFs and bookmarked papers from your library

My library

Recent searches

Try searching for:

Catch AI-generated content in scholarly and non-scholarly content

{ai} Detector

Ai Writer

Get PDF Summaries, highlighted text explanations 

Chat with PDF

Effortlessly create in-text citations and bibliographies in APA and 2,500 other formats

Citation generator

Get explanations, summaries, and answers on academic papers

Ease up your research workflow with {scispace}'s cohort of exciting AI tools

Elevate your academic writing skills and convey your ideas the way you want

Paraphraser

Explore our range of reading and writing tools

Your file is being prepared and should be ready in a few minutes. If it's a large file, it might take a bit longer. You can close this window, and we'll email you the file when it's done.

You have reached a maximum limit of <strong>{limit}</strong> columns in the table. Remove at least <strong>1</strong> column to add or create another one.

van Dalen

Author Tools

Chat about Author

Papers

Statistical models for noise-robust speech recognition