The use of temporal speech and lip information for multi-modal speaker identification via multi-stream HMMs

Investigates the use of temporal lip information, in conjunction with speech information, for robust, text-dependent speaker identification. We propose that significant speaker-dependent information can be obtained from moving lips, enabling speaker recognition systems to be highly robust in the presence of noise. The fusion structure for the audio and visual information is based around the use of multi-stream hidden Markov models (MSHMM), with audio and visual features forming two independent data streams. Recent work with multi-modal MSHMMs has been performed successfully for the task of speech recognition. The use of temporal lip information for speaker identification has been performed previously (T.J. Wark et al., 1998), however this has been restricted to output fusion via single-stream HMMs. We present an extension to this previous work, and show that a MSHMM is a valid structure for multi-modal speaker identification

Acoustic hazard detection for pedestrians with obscured hearing

Pedestrians’ use of mp3 players or mobile phones can pose the risk of being hit by motor vehicles. We present an approach for detecting a crash risk level using the computing power and the microphone of mobile devices that can be used to alert the user in advance of an approaching vehicle so as to avoid a crash. A single feature extractor classifier is not usually able to deal with the diversity of risky acoustic scenarios. In this paper, we address the problem of detection of vehicles approaching a pedestrian by a novel, simple, non resource intensive acoustic method. The method uses a set of existing statistical tools to mine signal features. Audio features are adaptively thresholded for relevance and classified with a three component heuristic. The resulting Acoustic Hazard Detection (AHD) system has a very low false positive detection rate. The results of this study could help mobile device manufacturers to embed the presented features into future potable devices and contribute to road safety.

Detection of rain in acoustic recordings of the environment

Environmental monitoring has become increasingly important due to the significant impact of human activities and climate change on biodiversity. Environmental sound sources such as rain and insect vocalizations are a rich and underexploited source of information in environmental audio recordings. This paper is concerned with the classification of rain within acoustic sensor re-cordings. We present the novel application of a set of features for classifying environmental acoustics: acoustic entropy, the acoustic complexity index, spectral cover, and background noise. In order to improve the performance of the rain classification system we automatically classify segments of environmental recordings into the classes of heavy rain or non-rain. A decision tree classifier is experientially compared with other classifiers. The experimental results show that our system is effective in classifying segments of environmental audio recordings with an accuracy of 93% for the binary classification of heavy rain/non-rain.

A Novel Multimodal Interface for Improving Visually Impaired People's Web Accessibility

This paper introduces a novel interface designed to help blind and visually impaired people to explore and navigate on the Web. In contrast to traditionally used assistive tools, such as screen readers and magnifiers, the new interface employs a combination of both audio and haptic features to provide spatial and navigational information to users. The haptic features are presented via a low-cost force feedback mouse allowing blind people to interact with the Web, in a similar fashion to their sighted counterparts. The audio provides navigational and textual information through the use of non-speech sounds and synthesised speech. Interacting with the multimodal interface offers a novel experience to target users, especially to those with total blindness. A series of experiments have been conducted to ascertain the usability of the interface and compare its performance to that of a traditional screen reader. Results have shown the advantages that the new multimodal interface offers blind and visually impaired people. This includes the enhanced perception of the spatial layout of Web pages, and navigation towards elements on a page. Certain issues regarding the design of the haptic and audio features raised in the evaluation are discussed and presented in terms of recommendations for future work.

Algorithmes de recommandation musicale

Ce mémoire est composé de trois articles qui s’unissent sous le thème de la recommandation musicale à grande échelle. Nous présentons d’abord une méthode pour effectuer des recommandations musicales en récoltant des étiquettes (tags) décrivant les items et en utilisant cette aura textuelle pour déterminer leur similarité. En plus d’effectuer des recommandations qui sont transparentes et personnalisables, notre méthode, basée sur le contenu, n’est pas victime des problèmes dont souffrent les systèmes de filtrage collaboratif, comme le problème du démarrage à froid (cold start problem). Nous présentons ensuite un algorithme d’apprentissage automatique qui applique des étiquettes à des chansons à partir d’attributs extraits de leur fichier audio. L’ensemble de données que nous utilisons est construit à partir d’une très grande quantité de données sociales provenant du site Last.fm. Nous présentons finalement un algorithme de génération automatique de liste d’écoute personnalisable qui apprend un espace de similarité musical à partir d’attributs audio extraits de chansons jouées dans des listes d’écoute de stations de radio commerciale. En plus d’utiliser cet espace de similarité, notre système prend aussi en compte un nuage d’étiquettes que l’utilisateur est en mesure de manipuler, ce qui lui permet de décrire de manière abstraite la sorte de musique qu’il désire écouter.

Espaces de timbre générés par des réseaux profonds convolutionnels

Il est avant-tout question, dans ce mémoire, de la modélisation du timbre grâce à des algorithmes d'apprentissage machine. Plus précisément, nous avons essayé de construire un espace de timbre en extrayant des caractéristiques du son à l'aide de machines de Boltzmann convolutionnelles profondes. Nous présentons d'abord un survol de l'apprentissage machine, avec emphase sur les machines de Boltzmann convolutionelles ainsi que les modèles dont elles sont dérivées. Nous présentons aussi un aperçu de la littérature concernant les espaces de timbre, et mettons en évidence quelque-unes de leurs limitations, dont le nombre limité de sons utilisés pour les construire. Pour pallier à ce problème, nous avons mis en place un outil nous permettant de générer des sons à volonté. Le système utilise à sa base des plug-ins qu'on peut combiner et dont on peut changer les paramètres pour créer une gamme virtuellement infinie de sons. Nous l'utilisons pour créer une gigantesque base de donnée de timbres générés aléatoirement constituée de vrais instruments et d'instruments synthétiques. Nous entrainons ensuite les machines de Boltzmann convolutionnelles profondes de façon non-supervisée sur ces timbres, et utilisons l'espace des caractéristiques produites comme espace de timbre. L'espace de timbre ainsi obtenu est meilleur qu'un espace semblable construit à l'aide de MFCC. Il est meilleur dans le sens où la distance entre deux timbres dans cet espace est plus semblable à celle perçue par un humain. Cependant, nous sommes encore loin d'atteindre les mêmes capacités qu'un humain. Nous proposons d'ailleurs quelques pistes d'amélioration pour s'en approcher.

DIMUSE: An integrated framework for distributed multimedia system with database management and security support

With the recent explosion in the complexity and amount of digital multimedia data, there has been a huge impact on the operations of various organizations in distinct areas, such as government services, education, medical care, business, entertainment, etc. To satisfy the growing demand of multimedia data management systems, an integrated framework called DIMUSE is proposed and deployed for distributed multimedia applications to offer a full scope of multimedia related tools and provide appealing experiences for the users. This research mainly focuses on video database modeling and retrieval by addressing a set of core challenges. First, a comprehensive multimedia database modeling mechanism called Hierarchical Markov Model Mediator (HMMM) is proposed to model high dimensional media data including video objects, low-level visual/audio features, as well as historical access patterns and frequencies. The associated retrieval and ranking algorithms are designed to support not only the general queries, but also the complicated temporal event pattern queries. Second, system training and learning methodologies are incorporated such that user interests are mined efficiently to improve the retrieval performance. Third, video clustering techniques are proposed to continuously increase the searching speed and accuracy by architecting a more efficient multimedia database structure. A distributed video management and retrieval system is designed and implemented to demonstrate the overall performance. The proposed approach is further customized for a mobile-based video retrieval system to solve the perception subjectivity issue by considering individual user's profile. Moreover, to deal with security and privacy issues and concerns in distributed multimedia applications, DIMUSE also incorporates a practical framework called SMARXO, which supports multilevel multimedia security control. SMARXO efficiently combines role-based access control (RBAC), XML and object-relational database management system (ORDBMS) to achieve the target of proficient security control. A distributed multimedia management system named DMMManager (Distributed MultiMedia Manager) is developed with the proposed framework DEMUR; to support multimedia capturing, analysis, retrieval, authoring and presentation in one single framework.

Dynamic visual features for audio-visual speaker verification

The cascading appearance-based (CAB) feature extraction technique has established itself as the state-of-the-art in extracting dynamic visual speech features for speech recognition. In this paper, we will focus on investigating the effectiveness of this technique for the related speaker verification application. By investigating the speaker verification ability of each stage of the cascade we will demonstrate that the same steps taken to reduce static speaker and environmental information for the visual speech recognition application also provide similar improvements for visual speaker recognition. A further study is conducted comparing synchronous HMM (SHMM) based fusion of CAB visual features and traditional perceptual linear predictive (PLP) acoustic features to show that higher complexity inherit in the SHMM approach does not appear to provide any improvement in the final audio-visual speaker verification system over simpler utterance level score fusion.

Vocal Tract Length Normalization Features for Audio Search

This paper presents speaker normalization approaches for audio search task. Conventional state-of-the-art feature set, viz., Mel Frequency Cepstral Coefficients (MFCC) is known to contain speaker-specific and linguistic information implicitly. This might create problem for speaker-independent audio search task. In this paper, universal warping-based approach is used for vocal tract length normalization in audio search. In particular, features such as scale transform and warped linear prediction are used to compensate speaker variability in audio matching. The advantage of these features over conventional feature set is that they apply universal frequency warping for both the templates to be matched during audio search. The performance of Scale Transform Cepstral Coefficients (STCC) and Warped Linear Prediction Cepstral Coefficients (WLPCC) are about 3% higher than the state-of-the-art MFCC feature sets on TIMIT database.

Cascading appearance-based features for visual voice activity detection

The detection of voice activity is a challenging problem, especially when the level of acoustic noise is high. Most current approaches only utilise the audio signal, making them susceptible to acoustic noise. An obvious approach to overcome this is to use the visual modality. The current state-of-the-art visual feature extraction technique is one that uses a cascade of visual features (i.e. 2D-DCT, feature mean normalisation, interstep LDA). In this paper, we investigate the effectiveness of this technique for the task of visual voice activity detection (VAD), and analyse each stage of the cascade and quantify the relative improvement in performance gained by each successive stage. The experiments were conducted on the CUAVE database and our results highlight that the dynamics of the visual modality can be used to good effect to improve visual voice activity detection performance.

Can audio-visual speech recognition outperform acoustically enhanced speech recognition in automotive environment?

The use of visual features in the form of lip movements to improve the performance of acoustic speech recognition has been shown to work well, particularly in noisy acoustic conditions. However, whether this technique can outperform speech recognition incorporating well-known acoustic enhancement techniques, such as spectral subtraction, or multi-channel beamforming is not known. This is an important question to be answered especially in an automotive environment, for the design of an efficient human-vehicle computer interface. We perform a variety of speech recognition experiments on a challenging automotive speech dataset and results show that synchronous HMM-based audio-visual fusion can outperform traditional single as well as multi-channel acoustic speech enhancement techniques. We also show that further improvement in recognition performance can be obtained by fusing speech-enhanced audio with the visual modality, demonstrating the complementary nature of the two robust speech recognition approaches.

Complete-linkage clustering for voice activity detection in audio and visual speech

We propose a novel technique for conducting robust voice activity detection (VAD) in high-noise recordings. We use Gaussian mixture modeling (GMM) to train two generic models; speech and non-speech. We then score smaller segments of a given (unseen) recording against each of these GMMs to obtain two respective likelihood scores for each segment. These scores are used to compute a dissimilarity measure between pairs of segments and to carry out complete-linkage clustering of the segments into speech and non-speech clusters. We compare the accuracy of our method against state-of-the-art and standardised VAD techniques to demonstrate an absolute improvement of 15% in half-total error rate (HTER) over the best performing baseline system and across the QUT-NOISE-TIMIT database. We then apply our approach to the Audio-Visual Database of American English (AVDBAE) to demonstrate the performance of our algorithm in using visual, audio-visual or a proposed fusion of these features.

Similarity-based birdcall retrieval from environmental audio

Automated digital recordings are useful for large-scale temporal and spatial environmental monitoring. An important research effort has been the automated classification of calling bird species. In this paper we examine a related task, retrieval of birdcalls from a database of audio recordings, similar to a user supplied query call. Such a retrieval task can sometimes be more useful than an automated classifier. We compare three approaches to similarity-based birdcall retrieval using spectral ridge features and two kinds of gradient features, structure tensor and the histogram of oriented gradients. The retrieval accuracy of our spectral ridge method is 94% compared to 82% for the structure tensor method and 90% for the histogram of gradients method. Additionally, this approach potentially offers a more compact representation and is more computationally efficient.

Generalised features for bird vocalisation retrieval in acoustic recordings

Bioacoustic monitoring has become a significant research topic for species diversity conservation. Due to the development of sensing techniques, acoustic sensors are widely deployed in the field to record animal sounds over a large spatial and temporal scale. With large volumes of collected audio data, it is essential to develop semi-automatic or automatic techniques to analyse the data. This can help ecologists make decisions on how to protect and promote the species diversity. This paper presents generic features to characterize a range of bird species for vocalisation retrieval. In the implementation, audio recordings are first converted to spectrograms using short-time Fourier transform, then a ridge detection method is applied to the spectrogram for detecting points of interest. Based on the detected points, a new region representation are explored for describing various bird vocalisations and a local descriptor including temporal entropy, frequency bin entropy and histogram of counts of four ridge directions is calculated for each sub-region. To speed up the retrieval process, indexing is carried out and the retrieved results are ranked according to similarity scores. The experiment results show that our proposed feature set can achieve 0.71 in term of retrieval success rate which outperforms spectral ridge features alone (0.55) and Mel frequency cepstral coefficients (0.36).

Acoustic classification of Australian anurans using syllable features

Acoustic classification of anurans (frogs) has received increasing attention for its promising application in biological and environment studies. In this study, a novel feature extraction method for frog call classification is presented based on the analysis of spectrograms. The frog calls are first automatically segmented into syllables. Then, spectral peak tracks are extracted to separate desired signal (frog calls) from background noise. The spectral peak tracks are used to extract various syllable features, including: syllable duration, dominant frequency, oscillation rate, frequency modulation, and energy modulation. Finally, a k-nearest neighbor classifier is used for classifying frog calls based on the results of principal component analysis. The experiment results show that syllable features can achieve an average classification accuracy of 90.5% which outperforms Mel-frequency cepstral coefficients features (79.0%).