Elmar Noeth

Friedrich-Alexander-Universität Erlangen-Nürnberg, Computer Science, Faculty Member

Followers

Following

Co-authors

Public Views

Interests

Uploads

Papers by Elmar Noeth

The prediction of focus

First European Conference on Speech Communication and Technology (Eurospeech 1989)

Sprachenunabhängige Verständlichkeitsanalyse bei Kindern mit orofazialen Spaltfehlbildungen auf Deutsch und Italienisch mittels akustischer Modellierung

Einleitung Zur objektiv-apparativen Bewertung der Verständlichkeit von Kindern mit Lippen-Kiefer-... more Einleitung Zur objektiv-apparativen Bewertung der Verständlichkeit von Kindern mit Lippen-Kiefer-Gaumenspalten (LKG) werden von der berichtenden Arbeitsgruppe für mehrere Sprachen erfolgreich Spracherkennungssysteme eingesetzt [ref:1]. Das Training eines Spracherkenners für[for full text, please go to the a.m. URL]

Download

Automatische Verständlichkeitsbewertung von Telefonaufnahmen Larynxteilresezierter mittels prosodischer Analyse

Einleitung Objektiv-apparative Stimmbewertungen werden derzeit meist auf der Basis gehaltener Vok... more Einleitung Objektiv-apparative Stimmbewertungen werden derzeit meist auf der Basis gehaltener Vokale durchgeführt. Jedoch reflektiert ein isolierter Vokal keine reale Kommunikationssituation. In früheren Arbeiten wurde gezeigt, dass prosodische Analyseverfahren verwendet werden können,[for full text, please go to the a.m. URL]

Download

Untersuchung von Sprachaufnahmen heiserer Stimmen mittels phonologischer und phonemischer Merkmale

Einleitung Gängige Methoden der automatischen Stimmanalyse stützen sich meist auf Aufnahmen gehal... more Einleitung Gängige Methoden der automatischen Stimmanalyse stützen sich meist auf Aufnahmen gehaltener Vokale. Sprachaufnahmen hingegen enthalten Stimmeinsätze, Schwankungen der Grundfrequenz F0 und auch Sprechpausen und erlauben so eine wesentlich umfangreichere Analyse. Neuere Entwicklungen[for full text, please go to the a.m. URL]

Verständlichkeitsbewertung von Telefonaufnahmen Larynxteilresezierter mittels der Kombination von automatischer Spracherkennung und prosodischer Analyse

Einleitung In früheren Arbeiten wurde gezeigt, dass prosodische Analyseverfahren verwendet werden... more Einleitung In früheren Arbeiten wurde gezeigt, dass prosodische Analyseverfahren verwendet werden können, um die Sprachverständlichkeit von pathologischen Sprechern automatisch zu bewerten [ref:1], [ref:2]. Der Fokus dieser Studie lag auf der Verständlichkeit[for full text, please go to the a.m. URL]

Download

Automatic evaluation of parkinson's speech — acoustic, prosodic and voice related cues

Interspeech 2013, 2013

Articulation and phonation is affected in 70 % to 90 % of patients with Parkinson's disease (PD).... more Articulation and phonation is affected in 70 % to 90 % of patients with Parkinson's disease (PD). This study focuses on the question whether speech carries information about 1. PD being present at a speaker or not, and 2. estimating the severity of PD (if present). We first perform classification experiments focusing on the automatic detection of PD as a 2-class problem (PD vs. healthy speakers). The detection of severity is described as a 3-class task based on the Unified Parkinson's Disease Rating Scale (UPDRS) ratings. We employ acoustic, prosodic and glottal features on different kinds of speech tests: various syllable repetition tasks, read sentences and texts, and monologues. Classification is performed in either case by SVMs. We report recognition results of 81.9 % when trying to differentiate between normally speaking persons and speakers with PD. With system fusion we achieved a recognition results of 59.1 % on the task of UPDRS classification.

Download

Verständlichkeitsanalyse bei Kindern mit Lippen-Kiefer-Gaumenspalten auf Deutsch und Italienisch mittels automatischer Spracherkennung

Die subjektive Bewertung von Verständlichkeit unterliegt Schwankungen zwischen verschiedenen Bewe... more Die subjektive Bewertung von Verständlichkeit unterliegt Schwankungen zwischen verschiedenen Bewertern, aber auch zwischen verschiedenen Bewertungssitzungen der gleichen Bewerter [ref:1]. Daher ist eine automatische Methode zur Messung der Verständlichkeit notwendig. Bisher konnte[for full text, please go to the a.m. URL]

Download

Vergleich der subjektiv-auditiven RBH-Beurteilung mit apparativer prosodischer Analyse und Irregularitätsberechnung

Einleitung In fruheren Arbeiten wurde gezeigt, dass prosodische Analyseverfahren verwendet werden... more Einleitung In fruheren Arbeiten wurde gezeigt, dass prosodische Analyseverfahren verwendet werden konnen, um Stimm- und Sprecheigenschaften automatisch zu bewerten [ref:1]. Der Fokus dieser Studie lag auf dem Vergleich einer Kombination von textbasierter prosodischer Analyse (Lehrstuhl[for full text, please go to the a.m. URL]

Roger", "Sorry", "I'm still listening" : dialog guiding signals in information retrieval dialogs

During any kind of information retrieval dialog the repetition of parts of information just given... more During any kind of information retrieval dialog the repetition of parts of information just given by the dialog partner can often be observed As these repetitions are usually ellip tic the intonation is very important for determining the speakers intention In this paper prototypically the times of day repeated by the customer in train table inquiry dialogs are investigated A scheme is developed for the o cers reactions depending on the intonation of these repetitions it has been integrated into our speech understanding and dialog system EVAR cf Gaussian classi ers were trained for distinguishing the dialog guiding sig nals con rmation question and feedback recognition rates of up to were obtained INTRODUCTION Dialog systems for information retrieval are potential applications for human machine communication In human human dialogs very often parts of the information just given by the speaker is repeated by the partner It can be observed e g in train table inquiries that the customer ...

Download

Towards the Automatic Evaluation of Dysarthric Speech

After a stroke, the quality of the speech in patients is often reduced. This is usually caused by... more After a stroke, the quality of the speech in patients is often reduced. This is usually caused by a deficit of the motor abilities of the vocal tract. The result is slurred speech. In the various patients, however, very different forms can appear. In the course of therapy, evaluation of the speech quality is required to determine the success of the treatment. At the moment, this assessment is performed only perceptually. This form of assessment is subject to strong intraand inter-individual variation. Therefore, an "objective" assessment is not guaranteed. In this study we present a rater-independent method for the speech disorders in dysarthria. We use methods of automatic speech recognition. The idea is to determine the speech intelligibility the main outcome parameter of speech automatically by an automatic speech recognizer. Between the criterion "intelligibility" and the recognition rate of the automatic system a correlation of -0.89 was obtained in a first ...

Download

Phonetic and prosodic analysis of speech

In order to cope with the problems of spontaneous speech (including, for example, hesitations and... more In order to cope with the problems of spontaneous speech (including, for example, hesitations and non-words) it is necessary to extract from the speech signal all information it contains. Modeling of words by segmental units should be supported by suprasegmental units since valuable information is represented in the prosody of an utterance. We present an approach to flexible and efficient modeling of speech by segmental units and describe extraction and use of suprasegmental information.

Download

Filled pauses in spontaneous speech

Filled pauses as e g uh eh signal dis uencies i e hesitations or repairs They do normally not occ... more Filled pauses as e g uh eh signal dis uencies i e hesitations or repairs They do normally not occur in read speech and were therefore up to now rather seldom investigated they must however be ac counted for in the automatic processing of spontaneous speech We present de scriptive statistics and the results of an au tomatic classi cation of lled pauses in the database of the VERBMOBIL project and discuss the relevancy of di erent prosodic features for the marking of di erent types

Download

Sprachverarbeitung jenseits der Spracherkennung

Multimedia — Informationssysteme zwischen Bild und Sprache, 1999

In diesem Beitrag wollen wir Ansatze zur Interpretation gesprochener Sprache prasentieren. Wir ze... more In diesem Beitrag wollen wir Ansatze zur Interpretation gesprochener Sprache prasentieren. Wir zeigen zunachst, warum die direkte Ubertragung der linguistischen Ansatze zur Interpretation von geschriebener Sprache auf gesprochene Sprache in der Regel nicht erfolgreich ist. Dies fuhrt uns zur Entwicklung eines Konzepts fur partielles Parsen, welches stochastische Verfahren und prosodische Information verwendet. Wir prasentieren vorlaufige Ergebnisse fur die einzelnen Wissensquellen, ein Test des Gesamtkonzepts steht noch aus.

Combining phonological and acoustic ASR-free features for pathological speech intelligibility assessment

Interspeech 2011, 2011

Automatic detection of hypernasal speech signals using nonlinear and entropy measurements

Interspeech 2012, 2012

Download

Automatic modelling of depressed speech: relevant features and relevance of gender

Interspeech 2014, 2014

Depression is an affective disorder characterised by psychomotor retardation; in speech, this sho... more Depression is an affective disorder characterised by psychomotor retardation; in speech, this shows up in reduction of pitch (variation, range), loudness, and tempo, and in voice qualities different from those of typical modal speech. A similar reduction can be observed in sleepy speech (relaxation). In this paper, we employ a small group of acoustic features modelling prosody and spectrum that have been proven successful in the modelling of sleepy speech, enriched with voice quality features, for the modelling of depressed speech within a regression approach. This knowledge-based approach is complemented by and compared with brute-forcing and automatic feature selection. We further discuss gender differences and the contributions of (groups of) features both for the modelling of depression and across depression and sleepiness.

Download

A language-independent feature set for the automatic evaluation of prosody

Interspeech 2009, 2009

From discourse to pathology: Automatic identification of Parkinson's disease patients via morphological measures across three languages

Cortex, 2020

Embodied cognition research on Parkinson's disease (PD) points to disruptions of frontostriatal l... more Embodied cognition research on Parkinson's disease (PD) points to disruptions of frontostriatal language functions as sensitive targets for clinical assessment. However, no existing approach has been tested for crosslinguistic validity, let alone by combining naturalistic tasks with machinelearning tools. To address these issues, we conducted the first classifier-based examination of morphological processing (a core frontostriatal function) in spontaneous monologues from PD patients across three typologically different languages. The study comprised 330 participants, encompassing speakers of Spanish (61 patients, 57 matched controls), German (88 patients, 88 matched controls), and Czech (20 patients, 16 matched controls). All subjects described the activities they perform during a regular day, and their monologues were automatically coded via morphological tagging, a computerized method that labels each word with a part-of-speech tag (e.g., noun, verb) and specific morphological tags (e.g., person, gender, number, tense). The ensuing data were subjected to machine-learning analyses to assess whether differential morphological patterns could classify between patients and controls and reflect the former's degree of motor impairment. Results showed robust classification rates, with over 80% of patients being discriminated from controls in each language separately. Moreover, the most discriminative morphological features were associated with the patients' motor compromise (as indicated by Pearson r correlations between predicted and collected motor impairment scores that ranged from moderate to moderate-to-strong across languages). Taken together, our results suggest that the morphological patterning, an embodied frontostriatal domain, may be distinctively affected in PD across languages and even under ecological testing conditions.

Download

A Multitask Learning Approach to Assess the Dysarthria Severity in Patients with Parkinson's Disease

Interspeech 2018, 2018

Parkinson's disease is a neurodegenerative disorder characterized by a variety of motor and non-m... more Parkinson's disease is a neurodegenerative disorder characterized by a variety of motor and non-motor symptoms. Particularly, several speech impairments appear in the initial stages of the disease, which affect aspects related to respiration and the movement of muscles and limbs in the vocal tract. Most of the studies in the literature aim to assess only one specific task from the patients, such as the classification of patients vs. healthy speakers, or the assessment of the neurological state of the patients. This study proposes a multitask learning approach based on convolutional neural networks to assess at the same time several speech deficits of the patients. A total of eleven speech aspects are considered, including difficulties of the patients to move articulators such as lips, palate, tongue, and larynx. According to the results, the proposed approach improves the generalization of the convolutional network, producing more representative feature maps to assess the different speech symptoms of the patients. The multitask learning scheme improves in of up to 4% the average accuracy relative to single networks trained to assess each individual speech aspect.

Download

Feature Space Visualization with Spatial Similarity Maps for Pathological Speech Data

Interspeech 2019, 2019

The feature vectors of a data set encode information about relations between speaker groups, clus... more The feature vectors of a data set encode information about relations between speaker groups, clusters and outliers. Based on the assumption that these relations are conserved within the spatial properties of feature vectors, we introduce similarity maps to visualize consistencies and deviations in magnitude and orientation between two feature vectors. We also present an iterative approach to find subspaces of a high-dimensional feature space that encode information about predefined speaker clusters. The methods were evaluated with two different data sets, one from chronically hoarse speakers and a second one from Parkinson's Disease patients and a healthy control group. The results showed that similarity maps provide a decent visualization of speaker groups and the spatial properties of their respective feature vectors. With the iterative optimization, it was possible to find features that show pronounced spatial differences between predefined clusters.

Download

Elmar Noeth

Uploads

Papers by Elmar Noeth

Log In