Up to 50% of delay in traffic is due to non-reoccurring events such as traffic accidents. Accidents lead to delays, which can be costly for transport companies. Road authorities are also very interested in warning drivers about accidents, e.g., to reroute them. This paper presents a novel and efficient approach and system for uncovering effects from traffic accidents by dynamic integration of GPS, weather, and traffic-accident data. This integration makes it possible to explore and quantify how accidents affects traffic. Dynamic integration means that data is combined at query time as it becomes available. This is necessary, because data can be missing (weather station down) or late arriving (accident not officially reported by the police yet). Further, the integration can be parameterized by the user, e.g., distance to accident, which is important due to inaccuracy in reporting. We present the integrated data on a map and show the effectiveness of the integration by allowing users to interactively browse all accidents or pick a single accident to study it in very fine-grained details. Using information from 31 433 road accidents and 38 billion GPS records, we show that the proposed dynamic data integration scales so very large data sets.

The focus on reducing CO2 emissions from the transport sector is larger than ever. Increasingly stricter reductions on fuel consumption and emissions are being introduced by the EU, e.g., to reduce the air pollution in many larger cities. Large sets of high-frequent GPS data from vehicles already exist. However, fuel consumption data is still rarely collected even though it is possible to measure the fuel consumption with high accuracy, e.g., using an OBD-II device and a smartphone. This paper, presents a method for comparing fuel-consumption estimates using the SIDRA TRIP model with real fuel measures to determine if the fuel-consumption model is sufficiently accurate. The model is implemented using a 2D, a simple 3D, and a high-precision (H3D) road map of Denmark. The original 2D map is lifted to a 3D map using a Digital Elevation Model (DEM). Results show that introducing a 3D map improves the accuracy of fuel-consumption estimates with up to 40% on hilly roads. There is only very little improvement of the high-precision (H3D) map over the simple 3D map. The fuel consumption estimates are most accurate on flat terrain with average fuel estimates of up to 99% accuracy. The fuel estimates are most inaccurate uphill/downhill and when the vehicles accelerate at speeds above 50 km/h.

GPS data is often used for computing travel time in road networks. In addition, GPS data is often map matched to find the routes driven by vehicles. Today GPS data is collected with different sampling periods, however, both the computed travel times and the routes found by map matching algorithms actual depends on the sampling period. This paper proposes a generic approach to study how travel time and map matched routes vary with the sampling period. Two types of map matching algorithms are used, point based where each position is handled individually, and trajectory based where positions from a vehicle is consider a data stream. A baseline is created using a real-world data set of 455 million positions from 368 vehicles collected with a sampling period of 1 second. This data set is downsampled to 8 data sets with sampling periods between 2 and 120 seconds. This downsampling enables an apple-to-apple comparison of travel time computation and route restoration for different sampling periods. The main conclusion is that travel times are reasonably accurate if the sampling period is 5 second or below for the point-based method and 20 seconds or below for the trajectory-based method. GPS data collected with 60 second is to inaccurate to be used for computing travel times. Trajectory-based map matching works best if the sampling period is 20 seconds or below.

Increasingly large volumes of vehicle trajectory data are becoming available. This data holds the potential to offer detailed insight into important aspects of vehicular transportation and road networks. This insight can in turn be utilized to enable a range of important services. Specifically, we demonstrate a system that is capable of leveraging very large collections of GPS trajectories for enabling interactive analyses of traffic in road intersections, which are often bottlenecks in road networks. These analyses are able to provide detailed insight into the time-varying functioning of intersections, and they offer a solid, data-driven foundation for improving the capacity of intersections and the overall road network. The system enables more cost-effective analyses than what is possible with traditional techniques. Demonstration participants will gain first-hand experience with interactive analyses on top of a database of some 40 billion GPS records capturing more than a billion km of driving.

The combined battery capacity in electric vehicles (EVs) is considered an integral part of balancing a smart power grid in the future. In addition, EVs can reduce the usage of fossil fuels in the transport sector because EVs can be charged using electricity from renewable energy sources, such as wind turbines. To both enable a smart grid and the use of renewable energy, it is essential to know when and where an EV is plugged into the power grid and what battery capacity is available. In this paper, we present a generic spatio-temporal data-warehouse model for storing detailed information on all aspects of charging EVs, including integration with the electricity prices from a spot market. The proposed data warehouse is fully implemented and currently contains 2.5 years of charging data from 176 EVs. We describe the date warehouse model and the implementation including complex operations such as spatially identifying charging station usage patterns. Further, we give examples of novel analyses, e.g., how the free battery capacity in the fleet of EVs changes over the day and how users can save money by charging the EVs when the electricity price is the lowest.

We propose a system for calculating the personalized annual fuel consumption and CO2 emissions from transportation. The system, named CO2NNIE, estimates the fuel consumption on the fastest route between the frequent destinations of the user. The travel time and fuel consumption estimated are based on 3.8 billion GPS records from 16 thousand cars and 198 million records from 218 cars annotated with fuel consumption data, respectively. The fuel consumption estimates from the system are validated using fuel-pump data. We find that estimates have good accuracy, i.e., are generally within 10% of the actual fuel consumption (4.6% deviation on average). We conclude, that the system provides new detailed information on CO2 emissions and fuel consumption for any make and model.

Reduction in greenhouse gas emissions from transportation attracts increasing interest from governments, fleet managers, and individual drivers. Eco-routing, which enables drivers to use eco-friendly routes, is a simple and effective approach to reducing emissions from transportation. We present EcoSky, a system that annotates edges of a road network with time dependent and uncertain eco-weights using GPS data and that supports different types of eco-routing. Basic eco-routing returns the most eco-friendly routes; skyline eco-routing takes into account not only fuel consumption but also travel time and distance when computing eco-routes; and personalized eco-routing considers each driver's past behavior and accordingly suggests different routes to different drivers.

GPS data recorded from driving vehicles is available from many sources and is a very good data foundation for answering traffic related queries. However, most approaches so far have not considered combining GPS data from many sources into a single data warehouse. Further, the integration of GPS data with fuel consumption data (from the so-called CAN bus in the vehicles) and weather data has not been done. In this paper, we propose a data warehouse design for handling GPS data, fuel consumption data, and weather data. The design is fully implemented in a running system using the PostgreSQL DBMS. The system has been in production since March 2011 and the main fact table contains today approximately 3.4 billion rows from 16 different data sources. We show that the system can be used for a number of novel traffic related analyses such as relating the fuel consumption of vehicles with the road network and road congestion.

Traffic researchers, planners, and analysts want a simple way to query the large quantities of GPS trajectories collected from vehicles. In addition, users expect the results to be presented immediately even when querying very large transportation networks with huge trajectory data sets. This paper presents a novel query type called sheaf, where users can browse trajectory data sets using a single mouse click. Sheaves are very versatile and can be used for location-based advertising, travel-time analysis, intersection analysis, and reachability analysis (isochrones). A novel in-memory trajectory index compresses the data by a factor of 12.4 and enables execution of sheaf queries in 40 ms. This is up to 2 orders of magnitude faster than existing work. We demonstrate the simplicity, versatility, and efficiency of sheaf queries using a real-world trajectory set consisting of 2.7 million trajectories (1.36 billion GPS records) and a network with 1.5 million edges.

The electric vehicle (EV) is an interesting vehicle type that can reduce the dependence on fossil fuels, e.g., by using electricity from wind turbines. A significant disadvantage of EVs is a very limited range, typically less than 200 km. This paper compares EVs to conventional vehicles (CVs) for private transportation using two very large data sets. The EV data set is collected from 164 vehicles (126 million rows) and the CV data set from 447 vehicles (206 million rows). Both data sets are collected in Denmark throughout 2012, with a logging frequency of 1 Hz. By comparing the two data sets, we observe that EVs are significantly slower on motorways, faster in cities, and drive shorter distances compared to CVs.

The mathematical theory of closed form functions for calculating LSFs on the basis of generating functions is presented. Exploiting recurrence relationships in the series expansion of Chebyshev polynomials of the first kind makes it possible to bootstrap iterative LSF-search from a set of characteristic polynomial zeros. The theoretical analysis is based on decomposition of sequences into symmetric and anti-symmetric polynomials defined as a series expansion of reduced Chebyshev polynomials of the first kind. Two variants of closed form functions are presented — each characterised by using a recurrence relationship in Chebyshev polynomials. The first exploits the well known three terms recurrence relationships of Chebyshev polynomials. The second hitherto unused recurrence properties of Chebyshev coefficients defining a set of coefficients and zeros used for bootstrapping calculation of LSFs. The theory is tested using bootstrapped calculation of zeros and by evaluating the complexity of the closed form function. The results of the lower complexity calculations show that real axis zeros are within a given iteration tolerance when compared to results of a standard root-finder. Full Paper

We present the INTRA system for interactive path-based traffic analysis. The analyses are developed in collaboration with traffic researchers and provide novel insights into conditions such as congestion, travel-time, choice of route, and traffic-flow. INTRA supports interactive point-and-click analysis, due to a novel and efficient indexing structure. With the web-site daisy.aau.dk/its/spqdemo/we will demonstrate several analyses, using a very large real-world data set consisting of 1.9 billion GPS records (1.5 million trajectories) recorded from more than 13 000 vehicles, and touching most of the road network in Denmark.

Reduction in greenhouse gas emissions from transportation is essential in combating global warming and climate change. Eco-routing enables drivers to use the most eco-friendly routes and is effective in reducing vehicle emissions. The EcoTour system assigns eco-weights to a road network based on GPS and fuel consumption data collected from vehicles to enable ecorouting. Given an arbitrary source-destination pair in Denmark, EcoTour returns the shortest route, the fastest route, and the eco-route, along with statistics for the three routes. EcoTour also serves as a testbed for exploring advanced solutions to a range of challenges related to eco-routing.

In this paper, a complete platform used to compute travel times from GPS data is described. Two approaches to computing travel time are proposed one based on points and one based on trips. Overall both approaches give reasonable results compared to existing manual estimated travel times. However, the trip-based approach requires more GPS data and of a higher quality than the point-based approach. The platform has been completely implemented using open-source software. The main conclusion is that large quantity of GPS data can be managed, with a limited budget and that GPS data is a good source for estimating travel times, if enough data is available.

In the present paper, a method is proposed for adaptive estimation and tracking of roots of time-varying, complex, and univariate polynomials, e.g.z-transform polynomials that arise from finite signal sequences.The objective with the method is to alleviate the computational burden induced by factorization.The estimation is done by solving a set of linear equations; the number of equations equals the order of the polynomial.To avoid potential drifting of the estimations, it is proposed to verify with Aberth-Ehrlich's factorization method at given intervals.A numerical experiment supplements theory by estimating roots of time-varying polynomials of different order.As a function of order, the proposed method has a lower run time than Lindsey-Fox and computing eigenvalues of companion matrices.The estimations are quite accurate, but tend to drift slightly in response to increasing coefficient pertubation lengths.

In recent studies, a non-parametric speech waveform representation (rep.) based on zeros of the z-transform (ZZT) has been proposed. The ZZT rep. has successfully been applied in separating mixed phase signals, e.g. pitch-synchronously windowed speech, into min/max phase by using the unit circle as discriminant. As the ZZT rep. is obtained by factorization of the z-transform, relations to the complex cepstrum (CC) exist. The present paper interrelates the ZZT rep. with the CC via factorization of the z-transform, and demonstrates that unit circle discrimination of a ZZT rep. can be formulated as a CC based separation by causality. A numerical experiment supplements theory by separating a range of LF glottal flow waveforms into their opening and closing phase constituents. Further, randomized mixed phase sequences are separated. As the CC based separation also can be obtained via FFT it has a lower time and space complexity than the ZZT based counterpart.

Current research has proposed a non-parametric speech waveform representation (rep) based on zeros of the z-transform (ZZT) [1] [2]. Empirically, the ZZT rep has successfully been applied in discriminating the glottal and vocal tract components in pitch-synchronously windowed speech by using the unit cir-cle (UC) as discriminant [1] [2]. Further, similarity between ZZT reps of windowed speech, glottal flow waveforms, and waveforms of glottal flow opening and closing phases has been demonstrated [1] [3]. Therefore, the underlying cause of the separation on either side of the UC can be analyzed via the individual ZZT reps of the opening and closing phase wave-forms; the waveforms are generated by the LF glottal flow model (GFM) [1]. The present paper demonstrates this cause and effect analytically and thereby supplement the previous em-pirical works. Moreover, this paper demonstrates that immisci-bility is variant under changes in frame lengths; lengths that maximize or minimize immiscibility are presented. Index Terms: Zeros of the z-transform, LF glottal flow model, opening/closing phase separation

Approximately 30% of the Danish population has severe problems in reading everyday text. In the light of the increasing amount of text available on the Internet this poses a democratic challenge to ensure “equal access” to information. The Talking Internet service - Access For All (AFA) - offers a free Internet-based tool for reading aloud any marked text with a synthetic voice. The only requirements are a standard equipped PC running a recent Windows OS and an Internet connection. Experiences gathered from running the service for more than 28 months underline the viability of the concept. There is a clear need for a free internet based Danish text-to-speech synthesizer. Furthermore, the current state of technology i.e. internet bandwidth, response time and server technology is sufficient for setting up an online automatic reading service that is used by steadily growing number of individuals and institutions.

A multi-stream speech recogniser is based on the combination of multiple feature streams each containing complementary information. In the past, multi-stream research has typically focused on systems that use a single feature extraction method. This heritage from conventional speech recognisers is an unnecessary restriction and both psychoacoustic and phonetic knowledge strongly motivate the use of heterogeneous features. In this paper we investigate how heterogeneous processing can be used in two different multi-stream configurations: first, a system where each stream handles a different frequency region of the speech (a multi-band recogniser) and, second a multi-stream recogniser where each stream handles the full frequency region. For each type of system we compare the performance using both homogeneous and heterogeneous processing. We demonstrate that the use of heterogeneous information significantly improves the clean speech recognition performance motivating us to continue exploring more specifically designed stream processing.

Two systems (statistical trajectory models (STM) and continuous density HMMs) utilizing three preprocessing methodologies (MFCC, RASTA and FBDYN) were evaluated on two databases, namely CTIMIT and the corresponding down-sampled TIMIT. Within the bounds of the experimental setup the comparative performance analysis showed that the STM significantly outperforms the HMM system on the CTIMIT database. Specifically, the performance of the STM system was found to be at least 10% better as compared to the one obtained by HMM when the RASTA preprocessing was used. The performance of both systems with FBDYN parametrization was found to be inferior to those using MFCC and RASTA. On the other hand, in low-noise conditions on the TIMIT database FBDYN yielded an improved performance for the HMM system, whereas STM achieved the best results with the MFCC parametrization.

decoding, the second transforms the parameters from This work is concerned with the subject of language- the decoding module and classifies the language. identification (LID). Two central issues are addressed. The common acoustic signal preprocessor calculates The first is to analyse the trade-off between detailed 12 RASTA filtered MFCC's, their first derivatives and acoustic modelling and robust estimation of acoustic the delta-log-energy. The phone and language decoding and language models. The second to find the optimal module consists of three parallel branches. In each of combination of acoustic and language scores for language- these the phone recogniser matches the acoustic identification. parameters to the acoustic models used by that recogniser. Experiments are carried out using the three languages The output from each recogniser is further matched American-English, German and Spanish from the OGI-TS against three language models. database. It is shown that on the average th...

Recently, we described a two-step self-learning approach for grapheme-to-phoneme (G2P) conversion [1].In the first step, grapheme and phoneme strings in the training data are aligned via an iterative Viterbi procedure that may insert graphemic and phonemic nulls where required.In the second step, a Trie structure encoding pronunciation rules is generated.In this paper we describe the alignment module, and give alignment accuracies on the NETtalk database.We also compare transcription accuracies for two approaches to the second step on three databases: the NETtalk database, the CMU dictionary and the French part of the ONOMASTICA lexicon.The two transcription approaches applied in this research are a Trie approach [1] and an approach based on binary decision trees grown by means of the Gelfand-Ravishankar-Delp algorithm [2,3,4].We discuss the choice of questions for these decision trees -it may be possible to formulate questions about groups of characters (e.g., "is the next letter a vowel?") that yield better trees than those that only use questions about individual characters (e.g., "is the next letter an 'A' ?").Finally, we discuss the implications of our work for G2P conversion.

A Self-Learning system for Grapheme to Phoneme conversion is described and tested. The system acquires the knowledge needed for grapheme-to-phoneme conversion from a training session in which a large number of pairs of grapheme strings and their corresponding (manually verified) phonemic transcription strings are presented to the system. The result from the training is a stochastic decision tree in which statistics - as given in the training material - about corresponding graphemes and phonemes are stored for later retrieval. The system is tested on a number of European languages and results from three tests are reported. In the first test, which concerns proper names, only the most probable phoneme candidate at each leaf of the tree is utilised. The second and the third test, both using a database of ordinary words, aims at analysing phoneme and word accuracies resulting from using N-Best phonemes at each leaf and from introducing phonotactic information, respectively. Using N-Best ...

The research reported in this paper presents a method to identify poly- and mono-phonemes for four European languages. The functionality of the poly-phonemes is tested in two experiments, and a limited set of mono-phonemes is identified for a language-identification experiment. Ten acoustically-similar speech sounds were identified across the four languages British-English, Danish, German, and Italian. These sounds, which constitute a substantial proportion of the phonemes of each language, are designated as (language independent) poly-phonemes, and may serve as a multi-lingual training base for labelling and recognition systems. The remaining sounds of each language, which do not fulfil the similarity conditions, are dubbed mono-phonemes. Two application experiments were conducted. In the first the poly-phonemes are applied in a label alignment task. In the second a small selected of mono-phonemes for each of the four languages is used in a preliminary test of the ability of these sets to serve as language discriminators.>

Results are reported from research on the use of continuously valued acoustic-phonetic features in the multi-language label alignment of combined speech corpora from three European languages: Danish, English, and Italian. A self-organizing neural network is used to transform cepstrum coefficients into a set of features, which are subsequently transformed into a set of principal components. These are used to model individual phonemes, which are used in a Viterbi search/level-building process to align an independently given string of phonemes with the corresponding speech signal. The results obtained show an overall accuracy of 55.7% in the positioning of the label boundary transitions in the combined test corpus. Detailed analysis shows that certain sound class boundaries are very accurately positioned. To provide a general solution to the problem of positioning badly positioned boundaries, an interactive component of the alignment system has been developed. First results demonstrate this component to be very valuable in the task of user-assisted boundary positioning.>

In previous work on label alignment, encouraging results were obtained using selected acoustic-phonetic features to model the individuals speech phonemes. Selection was based on minimal covariance between features on the one hand, and the inclusion of features underlying critical phonological opposition on the other. In the present work, principal component analysis was applied to give a number of uncorrelated output parameters which maximally exploit the discriminatory power of the features and are derived independently of the phonological functionality. Results of label alignment on three different European languages, Danish, English, and Italian, using different numbers of principal parameters show that the accuracy with ten parameters is at least as good as with 15 manually selected features. The best result is found for British English, which has 78% of its phoneme transition boundaries positioned within +or-20 ms of manually placed reference boundaries.>