A Likelihood-Maximizing Framework for Enhanced In-Car Speech Recognition Based on Speech Dialog System Interaction

Speech recognition in car environments has been identified as a valuable means for reducing driver distraction when operating noncritical in-car systems. Under such conditions, however, speech recognition accuracy degrades significantly, and techniques such as speech enhancement are required to improve these accuracies. Likelihood-maximizing (LIMA) frameworks optimize speech enhancement algorithms based on recognized state sequences rather than traditional signal-level criteria such as maximizing signal-to-noise ratio. LIMA frameworks typically require calibration utterances to generate optimized enhancement parameters that are used for all subsequent utterances. Under such a scheme, suboptimal recognition performance occurs in noise conditions that are significantly different from that present during the calibration session – a serious problem in rapidly changing noise environments out on the open road. In this chapter, we propose a dialog-based design that allows regular optimization iterations in order to track the ever-changing noise conditions. Experiments using Mel-filterbank noise subtraction (MFNS) are performed to determine the optimization requirements for vehicular environments and show that minimal optimization is required to improve speech recognition, avoid over-optimization, and ultimately assist with semireal-time operation. It is also shown that the proposed design is able to provide improved recognition performance over frameworks incorporating a calibration session only.

A communication model and partitioning algorithm for streaming applications for an embedded MPSoC

Energy efficient embedded computing enables new application scenarios in mobile devices like software-defined radio and video processing. The hierarchical multiprocessor considered in this work may contain dozens or hundreds of resource efficient VLIW CPUs. Programming this number of CPU cores is a complex task requiring compiler support. The stream programming paradigm provides beneficial properties that help to support automatic partitioning. This work describes a compiler for streaming applications targeting the self-build hierarchical CoreVA-MPSoC multiprocessor platform. The compiler is supported by a programming model that is tailored to fit the streaming programming paradigm. We present a novel simulated-annealing (SA) based partitioning algorithm, called Smart SA. The overall speedup of Smart SA is 12.84 for an MPSoC with 16 CPU cores compared to a single CPU implementation. Comparison with a state of the art partitioning algorithm shows an average performance improvement of 34.07%.

Automated fit quantification of tibial nail designs during the insertion using computer three-dimensional modelling

Intramedullary nailing is the standard fixation method for displaced diaphyseal fractures of the tibia. An optimal nail design should both facilitate insertion and anatomically fit the bone geometry at its final position in order to reduce the risk of stress fractures and malalignments. Due to the nonexistence of suitable commercial software, we developed a software tool for the automated fit assessment of nail designs. Furthermore, we demonstrated that an optimised nail, which fits better at the final position, is also easier to insert. Three-dimensional models of two nail designs and 20 tibiae were used. The fitting was quantified in terms of surface area, maximum distance, sum of surface areas and sum of maximum distances by which the nail was protruding into the cortex. The software was programmed to insert the nail into the bone model and to quantify the fit at defined increment levels. On average, the misfit during the insertion in terms of the four fitting parameters was smaller for the Expert Tibial Nail Proximal bend (476.3 mm2, 1.5 mm, 2029.8 mm2, 6.5 mm) than the Expert Tibial Nail (736.7 mm2, 2.2 mm, 2491.4 mm2, 8.0 mm). The differences were statistically significant (p ≤ 0.05). The software could be used by nail implant manufacturers for the purpose of implant design validation.

Model-based analysis and optimization of bioreactor for hematopoietic stem cell cultivation

One of the problems to be solved in attaining the full potentials of hematopoietic stem cell (HSC) applications is the limited availability of the cells. Growing HSCs in a bioreactor offers an alternative solution to this problem. Besides, it also offers the advantages of eliminating labour intensive process as well as the possible contamination involved in the periodic nutrient replenishments in the traditional T-flask stem cell cultivation. In spite of this, the optimization of HSC cultivation in a bioreactor has been barely explored. This manuscript discusses the development of a mathematical model to describe the dynamics in nutrient distribution and cell concentration of an ex vivo HSC cultivation in a microchannel perfusion bioreactor. The model was further used to optimize the cultivation by proposing three alternative feeding strategies in order to prevent the occurrence of nutrient limitation in the bioreactor. The evaluation of these strategies, the periodic step change increase in the inlet oxygen concentration, the periodic step change increase in the media inflow, and the feedback control of media inflow, shows that these strategies can successfully improve the cell yield of the bioreactor. In general, the developed model is useful for the design and optimization of bioreactor operation.

Bio-inspired plume tracking algorithm for UAVS

There is an increased interest on the use of UAVs for environmental research and to track bush fire plumes, volcanic plumes or pollutant sources. The aim of this paper is to describe the theory and results of a bio-inspired plume tracking algorithm. A memory based and gradient based approach, were developed and compared. A method for generating sparse plumes was also developed. Results indicate the ability of the algorithms to track plumes in 2D and 3D.

Security Analysis of Randomize-Hash-then-Sign Digital Signatures

At CRYPTO 2006, Halevi and Krawczyk proposed two randomized hash function modes and analyzed the security of digital signature algorithms based on these constructions. They showed that the security of signature schemes based on the two randomized hash function modes relies on properties similar to the second preimage resistance rather than on the collision resistance property of the hash functions. One of the randomized hash function modes was named the RMX hash function mode and was recommended for practical purposes. The National Institute of Standards and Technology (NIST), USA standardized a variant of the RMX hash function mode and published this standard in the Special Publication (SP) 800-106. In this article, we first discuss a generic online birthday existential forgery attack of Dang and Perlner on the RMX-hash-then-sign schemes. We show that a variant of this attack can be applied to forge the other randomize-hash-then-sign schemes. We point out practical limitations of the generic forgery attack on the RMX-hash-then-sign schemes. We then show that these limitations can be overcome for the RMX-hash-then-sign schemes if it is easy to find fixed points for the underlying compression functions, such as for the Davies-Meyer construction used in the popular hash functions such as MD5 designed by Rivest and the SHA family of hash functions designed by the National Security Agency (NSA), USA and published by NIST in the Federal Information Processing Standards (FIPS). We show an online birthday forgery attack on this class of signatures by using a variant of Dean’s method of finding fixed point expandable messages for hash functions based on the Davies-Meyer construction. This forgery attack is also applicable to signature schemes based on the variant of RMX standardized by NIST in SP 800-106. We discuss some important applications of our attacks and discuss their applicability on signature schemes based on hash functions with ‘built-in’ randomization. Finally, we compare our attacks on randomize-hash-then-sign schemes with the generic forgery attacks on the standard hash-based message authentication code (HMAC).

Cryptanalysis of Tav-128 Hash Function

Many RFID protocols use cryptographic hash functions for their security. The resource constrained nature of RFID systems forces the use of light weight cryptographic algorithms. Tav-128 is one such 128-bit light weight hash function proposed by Peris-Lopez et al. for a low-cost RFID tag authentication protocol. Apart from some statistical tests for randomness by the designers themselves, Tav-128 has not undergone any other thorough security analysis. Based on these tests, the designers claimed that Tav-128 does not posses any trivial weaknesses. In this article, we carry out the first third party security analysis of Tav-128 and show that this hash function is neither collision resistant nor second preimage resistant. Firstly, we show a practical collision attack on Tav-128 having a complexity of 237 calls to the compression function and produce message pairs of arbitrary length which produce the same hash value under this hash function. We then show a second preimage attack on Tav-128 which succeeds with a complexity of 262 calls to the compression function. Finally, we study the constituent functions of Tav-128 and show that the concatenation of nonlinear functions A and B produces a 64-bit permutation from 32-bit messages. This could be a useful light weight primitive for future RFID protocols.

Side Channel Analysis of Some Hash Based MACs: A Response to SHA-3 Requirements

The forthcoming NIST’s Advanced Hash Standard (AHS) competition to select SHA-3 hash function requires that each candidate hash function submission must have at least one construction to support FIPS 198 HMAC application. As part of its evaluation, NIST is aiming to select either a candidate hash function which is more resistant to known side channel attacks (SCA) when plugged into HMAC, or that has an alternative MAC mode which is more resistant to known SCA than the other submitted alternatives. In response to this, we perform differential power analysis (DPA) on the possible smart card implementations of some of the recently proposed MAC alternatives to NMAC (a fully analyzed variant of HMAC) and HMAC algorithms and NMAC/HMAC versions of some recently proposed hash and compression function modes. We show that the recently proposed BNMAC and KMDP MAC schemes are even weaker than NMAC/HMAC against the DPA attacks, whereas multi-lane NMAC, EMD MAC and the keyed wide-pipe hash have similar security to NMAC against the DPA attacks. Our DPA attacks do not work on the NMAC setting of MDC-2, Grindahl and MAME compression functions.

Simultaneous underwater visibility assessment, enhancement and improved stereo

Vision-based underwater navigation and obstacle avoidance demands robust computer vision algorithms, particularly for operation in turbid water with reduced visibility. This paper describes a novel method for the simultaneous underwater image quality assessment, visibility enhancement and disparity computation to increase stereo range resolution under dynamic, natural lighting and turbid conditions. The technique estimates the visibility properties from a sparse 3D map of the original degraded image using a physical underwater light attenuation model. Firstly, an iterated distance-adaptive image contrast enhancement enables a dense disparity computation and visibility estimation. Secondly, using a light attenuation model for ocean water, a color corrected stereo underwater image is obtained along with a visibility distance estimate. Experimental results in shallow, naturally lit, high-turbidity coastal environments show the proposed technique improves range estimation over the original images as well as image quality and color for habitat classification. Furthermore, the recursiveness and robustness of the technique allows implementation onboard an Autonomous Underwater Vehicle for improving navigation and obstacle avoidance performance.

Automated sensory data alignment for environmental and epidermal change monitoring

In this paper we present research adapting a state of the art condition-invariant robotic place recognition algorithm to the role of automated inter- and intra-image alignment of sensor observations of environmental and skin change over time. The approach involves inverting the typical criteria placed upon navigation algorithms in robotics; we exploit rather than attempt to fix the limited camera viewpoint invariance of such algorithms, showing that approximate viewpoint repetition is realistic in a wide range of environments and medical applications. We demonstrate the algorithms automatically aligning challenging visual data from a range of real-world applications: ecological monitoring of environmental change, aerial observation of natural disasters including flooding, tsunamis and bushfires and tracking wound recovery and sun damage over time and present a prototype active guidance system for enforcing viewpoint repetition. We hope to provide an interesting case study for how traditional research criteria in robotics can be inverted to provide useful outcomes in applied situations.

A multidimensional collaborative filtering fusion approach with dimensionality reduction

Multidimensional data are getting increasing attention from researchers for creating better recommender systems in recent years. Additional metadata provides algorithms with more details for better understanding the interaction between users and items. While neighbourhood-based Collaborative Filtering (CF) approaches and latent factor models tackle this task in various ways effectively, they only utilize different partial structures of data. In this paper, we seek to delve into different types of relations in data and to understand the interaction between users and items more holistically. We propose a generic multidimensional CF fusion approach for top-N item recommendations. The proposed approach is capable of incorporating not only localized relations of user-user and item-item but also latent interaction between all dimensions of the data. Experimental results show significant improvements by the proposed approach in terms of recommendation accuracy.

Tensor reduction for user profiling in personalized recommender systems

User profiling is the process of constructing user models which represent personal characteristics and preferences of customers. User profiles play a central role in many recommender systems. Recommender systems recommend items to users based on user profiles, in which the items can be any objects which the users are interested in, such as documents, web pages, books, movies, etc. In recent years, multidimensional data are getting more and more attention for creating better recommender systems from both academia and industry. Additional metadata provides algorithms with more details for better understanding the interactions between users and items. However, most of the existing user/item profiling techniques for multidimensional data analyze data through splitting the multidimensional relations, which causes information loss of the multidimensionality. In this paper, we propose a user profiling approach using a tensor reduction algorithm, which we will show is based on a Tucker2 model. The proposed profiling approach incorporates latent interactions between all dimensions into user profiles, which significantly benefits the quality of neighborhood formation. We further propose to integrate the profiling approach into neighborhoodbased collaborative filtering recommender algorithms. Experimental results show significant improvements in terms of recommendation accuracy.

Sealing the leak on classical NTRU signatures

Initial attempts to obtain lattice based signatures were closely related to reducing a vector modulo the fundamental parallelepiped of a secret basis (like GGH [9], or NTRUSign [12]). This approach leaked some information on the secret, namely the shape of the parallelepiped, which has been exploited on practical attacks [24]. NTRUSign was an extremely efficient scheme, and thus there has been a noticeable interest on developing countermeasures to the attacks, but with little success [6]. In [8] Gentry, Peikert and Vaikuntanathan proposed a randomized version of Babai’s nearest plane algorithm such that the distribution of a reduced vector modulo a secret parallelepiped only depended on the size of the base used. Using this algorithm and generating large, close to uniform, public keys they managed to get provably secure GGH-like lattice-based signatures. Recently, Stehlé and Steinfeld obtained a provably secure scheme very close to NTRUSign [26] (from a theoretical point of view). In this paper we present an alternative approach to seal the leak of NTRUSign. Instead of modifying the lattices and algorithms used, we do a classic leaky NTRUSign signature and hide it with gaussian noise using techniques present in Lyubashevky’s signatures. Our main contributions are thus a set of strong NTRUSign parameters, obtained by taking into account latest known attacks against the scheme, a statistical way to hide the leaky NTRU signature so that this particular instantiation of CVP-based signature scheme becomes zero-knowledge and secure against forgeries, based on the worst-case hardness of the O~(N1.5)-Shortest Independent Vector Problem over NTRU lattices. Finally, we give a set of concrete parameters to gauge the efficiency of the obtained signature scheme.

A computer generated model of adenosine receptors rationalising binding and selectivity of receptor ligands

Computer graphic analyses on a broad spectrum of adenosine receptor ligands has shown that both the A1 and A2 adenosine receptors have three binding sites. The spatial relationship of these three binding sites has been defined. Adenosine orientation at A1 and A2 is different.