Search

This paper presents a context-dependent hybrid connectionist speech recognition system that uses a set of generalized hierarchical mixtures of experts (HME) to estimate context-dependent posterior acoustic class probabilities. The connectionist part of the system is organized in a modular fashion, allowing the distributed training of such a system on regular workstations. Context classes are based on polyphonic contexts, clustered using decision trees which we adopt from our continuous density HMM recognizer JANUS (Waibel et al., 1996). The system is evaluated on ESST, an English speaker-independent spontaneous speech database. Context dependent modeling is shown to yield significant improvements over simple context-independent modeling, requiring only small additional overhead in terms of training and decoding time.

Decoding algorithm is a crucial part in statistical machine translation. We describe a stack decoding algorithm in this paper. We present the hypothesis scoring method and the heuristics used in our algorithm. We report several techniques deployed to improve the performance of the decoder. We also introduce a simplified model to moderate the sparse data problem and to speed up the decoding process. We evaluate and compare these techniques/models in our statistical machine translation system.

The paper explores several statistical pattern recognition techniques to classify utterances according to their emotional content. The authors have recorded a corpus containing emotional speech with over a 1000 utterances from different speakers. They present a new method of extracting prosodic features from speech, based on a smoothing spline approximation of the pitch contour. To make maximal use of the limited amount of training data available, they introduce a novel pattern recognition technique: majority voting of subspace specialists. Using this technique, they obtain classification performance that is close to human performance on the task.

A collection of papers on styles of spoken communication. In this text, contributions on methodology are complemented by empirical studies exemplifying styles in different genres and activity-types in natural exchange.

This paper describes JANUS-III, our most recent version of the JANUS speech-to-speech translation system. We present an overview of the system and focus on how system design facilitates speech translation between multiple languages, and allows for easy adaptation to new source and target languages. We also describe our methodology for evaluation of end-to-end system performance with a variety of source and target languages. For system development and evaluation, we have experimented with both push-to-talk as well as cross-talk recording conditions. To date, our system has achieved performance levels of over 80% acceptable translations on transcribed input, and over 70% acceptable translations on speech input recognized with a 75-90% word accuracy. Our current major research is concentrated on enhancing the capabilities of the system to deal with input in broad and general domains.

This paper deals with buffer management for tree-based access-path structures in real-time database system kernels. A predictable and efficient buffer replacement strategy for B- and B/sup +/-trees is introduced. The strategy called LC (Level Control) uses the structural information known about the trees and the tree search to determine a node to be replaced. The basic idea is to use the level of a node as replacement criteria. This allows a certain prediction of buffer hits and misses for tree search operations, even if the item's value searched for is unknown. A small prediction interval can be calculated for search operations and a larger interval for update operations. Simulations show the correctness of these calculations and the efficiency of LC for B- and B/sup +/-trees compared to common other replacement strategies.

The paper describes one specific aspect of the software component construction in the life cycles of computer based systems. The construction is located following the requirements analysis, conception and design. Systems are designed as within the ECBS process so called services which consist of a set of objects working together, keeping the states of the system. To ensure an efficient and rapid construction of systems an easy to use mechanism to store and retrieve objects together with their relations is necessary. This demands an embedded method of keeping objects in a database-the persistence. The described mechanism is implemented using C++ and verified in some projects. A man machine service is used as an example to show the application of this approach.

Building robust stochastic language models is a major issue in speech recognition systems. Conventional word-based n-gram models do not capture any linguistic constraints inherent in speech. In this paper, the notion of function and content words (open/closed word classes) is used to provide linguistic knowledge that can be incorporated into language models. Function words are articles, prepositions and personal pronouns. Content words are nouns, verbs, adjectives and adverbs. Based on this class definition resulting in function and content word markers, a new language model is defined. A combination of the word-based model with this new model is introduced. The combined model shows modest improvements both in perplexity results and recognition performance.

This paper presents fill algorithms for boundary-defined regions in raster graphics. The algorithms require only a constant-size working memory. The methods presented are based on the so-called “seed fill” algorithms that use the internal connectivity of the region with a given inner point. Basic methods, as well as additional hcuristics for speeding up the algorithm, are described and verified. Empirical results are used to compare the time complexities of the algorithms for different classes of regions.

Inexpensive mobile data communication - based, e.g., on cellular phone networks - enables the access to data and services for "everyone, anytime, anywhere". At the same time, global information systems begin to emerge, which provide a unified access to multimedia data and services. However, the mobile access to these information repositories is being hampered by the technical constraints of the underlying hardware, especially with respect to interactive multimedia applications. Central challenges are low bandwidth in comparison to local-area networks, resource constraints of mobile devices, and massive changes in the information stock. Based on an integrative view of systems and data models for mobile information visualization, this paper discusses solution strategies which address these challenges. Furthermore, we discuss prototype applications being developed at our institute ("HyperFunk", "Active Multimedia Mail") and the basic research project "MOVI".

This paper reports the results of our experiments aimed at the automatic optimization of the number of parameters in the semi-continuous phonetically tied HMM based speech recognition system that is part of the speech-to-speech translation system JANUS-2. We propose different algorithms devised to determine the optimal number of model parameters. In recognition experiments performed on a spontaneous human-to-human dialog database, we show that automatic optimization of the acoustic modeling parameter size with the proposed algorithm improves the recognition performance without increasing the required amount of computing power and memory.

The simultaneous interpreter is a secondary text producer in the sense that s/he is not responsible for the contents of the text to be transferred to the target language. In this paper it is assumed 1) that this lack of autonomy results in the construction of a discourse model which differs from the speaker’s discourse model and 2) that this difference is reflected in different referential expressions used by the speaker and the interpreter respectively. Based on Mira Ariel’s Accessibility Theory, referential expressions are regarded as accessibility markers, by means of which the text producer signals to the hearer whether the mental representation of the referent in question has a high or low degree of accessibility in his/her memory. It is claimed that, whenever text production is difficult due to special conditions of the simultaneous mode, the interpreter tends to choose referential expressions which indicate that a particular referent has a lower state of activation in his/her discourse model as compared with the speaker’s. If a higher accessibility marker is used in such a case this can be explained as a misunderstanding, comprehension deficit or loss of information.

The management of complex spatial objects in applications, such as geography and cartography, imposes stringent new requirements on spatial database systems, in particular on efficient query processing. As shown before, the performance of spatial query processing can be improved by decomposing complex spatial objects into simple components. Up to now, only decomposition techniques generating a linear number of very simple components, e.g. triangles or trapezoids, have been considered. In this paper, we will investigate the natural trade-off between the complexity of the components and the redundancy, i.e. the number of components, with respect to its effect on efficient query processing. In particular, we present two new decomposition methods generating a better balance between the complexity and the number of components than previously known techniques. We compare these new decomposition methods to the traditional undecomposed representation as well as to the well-known decomposition into convex polygons with respect to their performance in spatial query processing. This comparison points out that for a wide range of query selectivity the new decomposition techniques clearly outperform both the undecomposed representation and the convex decomposition method. More important than the absolute gain in performance by a factor of up to an order of magnitude is the robust performance of our new decomposition techniques over the whole range of query selectivity.

The paper reports on an off-line programming and simulation system developed at the IPK-Berlin. Some relevant industrial applications are presented. Furthermore the integration of this basic components to a planning system for robotized workcells is described.

This work investigates protograph-based LDPC codes for the AWGN channel with OOK modulation. A non-uniform distribution of the OOK modulation symbols is considered to improve the power efficiency especially for low SNRs. To this end, a specific transmitter architecture based on time sharing is proposed that allows probabilistic shaping of (some) OOK modulation symbols. Tailored protograph-based LDPC code designs outperform standard schemes with uniform signaling and off-the-shelf codes by 1.1 dB for a transmission rate of 0.25 bits/channel use., Comment: Invited Paper for CISS 2019

In this paper, we present the first conditional generative adversarial network (cGAN) architecture that is specifically designed to fuse synthetic aperture radar (SAR) and optical multi-spectral (MS) image data to generate cloud- and haze-free MS optical data from a cloud-corrupted MS input and an auxiliary SAR image. Experiments on Sentinel-2 MS and Sentinel-l SAR data confirm that our extended SAR-Opt-cGAN model utilizes the auxiliary SAR information to better reconstruct MS images than an equivalent model which uses the same architecture but only single-sensor MS data as input.

In this paper we present our experience teaching Systems Program- ming in C to undergraduate students. Additionally to traditional Unix-like operating system approach, we employed a robotic plat- form - the e-puck mobile robot - to increase the students moti- vation and improve their learning experience. A robotic platform provides high attraction for students, making the class stand-out compared to other courses. Yet it is not only a playground, rather, the platform allows to present very challenging and sophisticated real-life programming problems in a tangible way. The chosen robot provides an open-source operating system with a well struc- tured programming interface and thus o ers a real-world, complex example of systems programming to the students. We describe the overall curriculum and the syllabus of the course itself. Emphasis is put on the design of the in-class and homework assignments, but the robotic platform is brie y described as well. Our success is con rmed by the end-of-semester evaluation by the students, who ranked our course among the top of all bachelor-level courses.