Publications
Autoregressive HMM speech synthesis
March 25, 2012
Conference Paper
Published in:
Proc. IEEE Int. Conf. on Acoustics, Speech and Signal Processing, ICASSP, 25-30 March 2012, pp. 4021-4.
Topic:
R&D area:
Summary
Autoregressive HMM modeling of spectral features has been proposed as a replacement for standard HMM speech synthesis. The merits of the approach are explored, and methods for enforcing stability of the estimated predictor coefficients are presented. It appears that rather than directly estimating autoregressive HMM parameters, greater synthesis accuracy is obtained by estimating the autoregressive HMM parameters by using a more traditional HMM recognition system to compute state-level posterior probabilities that are then used to accumulate statistics to estimate predictor coefficients. The result is a simplified mathematical framework that requires no modeling of derivatives and still provides smooth synthesis without unnatural spectral discontinuities. The resulting synthesis algorithm involves no matrix solves and may be formulated causally, and appears to result in quality very similar to that of more traditional HMM synthesis approaches. This paper describes the implementation of a complete Autoregressive HMM LVCSR system and its application for synthesis, and describes the preliminary synthesis results.
Summary
Autoregressive HMM modeling of spectral features has been proposed as a replacement for standard HMM speech synthesis. The merits of the approach are explored, and methods for enforcing stability of the estimated predictor coefficients are presented. It appears that rather than directly estimating autoregressive HMM parameters, greater synthesis accuracy is...
READ MORE
Goodness-of-fit statistics for anomaly detection in Chung-Lu random graphs
March 25, 2012
Conference Paper
Published in:
ICASSP 2012, Proc. IEEE Int. Conf. on Acoustics, Speech and Signal Processing, 25-30 March 2012, pp. 3265-8.
Topic:
R&D area:
Summary
Anomaly detection in graphs is a relevant problem in numerous applications. When determining whether an observation is anomalous with respect to the model of typical behavior, the notion of "goodness of fit" is important. This notion, however, is not well understood in the context of graph data. In this paper, we propose three goodness-of-fit statistics for Chung-Lu random graphs, and analyze their efficacy in discriminating graphs generated by the Chung-Lu model from those with anomalous topologies. In the results of a Monte Carlo simulation, we see that the most powerful statistic for anomaly detection depends on the type of anomaly, suggesting that a hybrid statistic would be the most powerful.
Summary
Anomaly detection in graphs is a relevant problem in numerous applications. When determining whether an observation is anomalous with respect to the model of typical behavior, the notion of "goodness of fit" is important. This notion, however, is not well understood in the context of graph data. In this paper...
READ MORE
Topic identification based extrinsic evaluation of summarization techniques applied to conversational speech
March 25, 2012
Journal Article
Published in:
Proc. IEEE Int. Conf. on Acoustics, Speech and Signal Processing, ICASSP, 25-30 March 2012, pp. 5073-6.
Topic:
R&D area:
Summary
Document summarization algorithms are most commonly evaluated according to the intrinsic quality of the summaries they produce. An alternate approach is to examine the extrinsic utility of a summary, measured by the ability of the summary to aid a human in the completion of a specific task. In this paper, we use topic identification as a proxy for relevancy determination in the context of an information retrieval task, and a summary is deemed effective if it enables a user to determine the topical content of a retrieved document. We utilize Amazon's Mechanical Turk service to perform a large-scale human study contrasting four different summarization systems applied to conversational speech from the Fisher Corpus. We show that these results appear to be correlated with the performance of an automated topic identification system, and argue that this automated system can act as a low-cost proxy for a human evaluation during the development stages of a summarization system.
Summary
Document summarization algorithms are most commonly evaluated according to the intrinsic quality of the summaries they produce. An alternate approach is to examine the extrinsic utility of a summary, measured by the ability of the summary to aid a human in the completion of a specific task. In this paper...
READ MORE
Moments of parameter estimates for Chung-Lu random graph models
March 25, 2012
Conference Paper
Published in:
ICASSP 2012, Proc. IEEE Int. Conf. on Acoustics, Speech, and Signal Processing, 25-30 March 2012, pp. 3961-4.
Topic:
R&D area:
R&D group:
Summary
As abstract representations of relational data, graphs and networks find wide use in a variety of fields, particularly when working in non- Euclidean spaces. Yet for graphs to be truly useful in in the context of signal processing, one ultimately must have access to flexible and tractable statistical models. One model currently in use is the Chung- Lu random graph model, in which edge probabilities are expressed in terms of a given expected degree sequence. An advantage of this model is that its parameters can be obtained via a simple, standard estimator. Although this estimator is used frequently, its statistical properties have not been fully studied. In this paper, we develop a central limit theory for a simplified version of the Chung-Lu parameter estimator. We then derive approximations for moments of the general estimator using the delta method, and confirm the effectiveness of these approximations through empirical examples.
Summary
As abstract representations of relational data, graphs and networks find wide use in a variety of fields, particularly when working in non- Euclidean spaces. Yet for graphs to be truly useful in in the context of signal processing, one ultimately must have access to flexible and tractable statistical models. One...
READ MORE
Dynamic Distributed Dimensional Data Model (D4M) database and computation system
March 25, 2012
Conference Paper
Published in:
ICASSP 2012, Proc. IEEE Int. Conf. on Acoustics, Speech and Signal Processing, 25-30 March 2012, pp. 5349-52.
Summary
A crucial element of large web companies is their ability to collect and analyze massive amounts of data. Tuple store databases are a key enabling technology employed by many of these companies (e.g., Google Big Table and Amazon Dynamo). Tuple stores are highly scalable and run on commodity clusters, but lack interfaces to support efficient development of mathematically based analytics. D4M (Dynamic Distributed Dimensional Data Model) has been developed to provide a mathematically rich interface to tuple stores (and structured query language "SQL" databases). D4M allows linear algebra to be readily applied to databases. Using D4M, it is possible to create composable analytics with significantly less effort than using traditional approaches. This work describes the D4M technology and its application and performance.
Summary
A crucial element of large web companies is their ability to collect and analyze massive amounts of data. Tuple store databases are a key enabling technology employed by many of these companies (e.g., Google Big Table and Amazon Dynamo). Tuple stores are highly scalable and run on commodity clusters, but...
READ MORE
Design and analysis of a hyperspectral microwave receiver subsystem
March 9, 2012
Conference Paper
Published in:
MICRORAD 2012, 12th Specialist Meeting on Microwave Radiometry and Remote Sensing of the Environment, 5-9 March 2012.
Topic:
R&D area:
Summary
Recent technology advances have profoundly changed the landscape of modern radiometry by enabling miniaturized, low-power, and low-noise radio-frequency receivers operating at frequencies near 200 GHz and beyond. These advances enable the practical use of receiver arrays to multiplex multiple broad frequency bands into many spectral channels. We use the term "hyperspectral microwave" to refer generically to microwave sounding systems with approximately 50 spectral channels or more. In this paper, we report on the design and analysis of the receiver subsystem (lensed antenna, RF frontend electronics, and IF processor module) for the Hyperspectral Microwave Atmospheric Sounder (HyMAS) comprising multiple receivers near the oxygen absorption line at 118.75 GHz and the water vapor absorption line at 183.31 GHz. The hyperspectral microwave receiver system will be integrated into a new scanhead compatible with the NASA GSFC Conical Scanning Microwave Imaging Radiometer/Compact Submillimeter-wave Imaging Radiometer (CoSMIR/CoSSIR) airborne instrument system to facilitate demonstration and performance characterization under funding from the NASA ESTO Advanced Component Technology program. Four identical radiometers will be used to cover 108-119 GHz, and two identical receivers will be used to cover 173-183 GHz. Subharmonic mixers will be driven by frequency-multiplied dielectric resonant oscillators, and single-sideband operation will be achieved by waveguide filtering of the lower sideband. A relatively high IF frequency is chosen to facilitate miniaturization of the IF processor module, which will be fabricated using Low Temperature Co-fired Ceramic (LTCC) technology. Corrugated feed antennas with lenses are used to achieve a FWHM beamwidth of approximately 3.5 degrees. Two polarizations are measured by each feed to increase overall channel count, and multiple options will be considered during the design phase for the polarization diplexing approach. Broadband operation over a relatively high intermediate frequency range (18-29 GHz) is a technical challenge of the front-end receiver systems, and a receiver temperature of approximately 2000-3000K is expected over the receiver bandwidth. This performance, together with approximately l00-msec integration times typical of airborne operation, yields channel NEDTs of approximately 0.35K, which is adequate to demonstrate the hyperspectral microwave concept by comparing profile retrievals with high-fidelity ground truth available either by coincident overpasses of hyperspectral infrared sounders and/or in situ radiosonde/dropsonde measurements.
Summary
Recent technology advances have profoundly changed the landscape of modern radiometry by enabling miniaturized, low-power, and low-noise radio-frequency receivers operating at frequencies near 200 GHz and beyond. These advances enable the practical use of receiver arrays to multiplex multiple broad frequency bands into many spectral channels. We use the term...
READ MORE
Hazard alerting based on probabilistic models
March 1, 2012
Journal Article
Published in:
J. Guidance, Control, Dynamics, Vol. 35, No. 2, March-April 2012, pp. 442-450.
Topic:
R&D area:
R&D group:
Summary
Hazard alerting systems alert operators to potential future undesirable events so that action may be taken to mitigate risk. One way to develop a hazard alerting system based on probabilistic models is by using a threshold-based approach, where the probability of the undesirable event without mitigation is compared against a threshold. Another way to develop such a system is to model the system as a Markov decision process and solve for the hazard experiments reveal that an expected utility approach performs better than threshold-based approaches when the dynamic stochasticity is high, where accounting for delays or changes in the alert becomes more important. however, for certain system parameters and operating environments, a threshold-based approach may provide comparable performance.
Summary
Hazard alerting systems alert operators to potential future undesirable events so that action may be taken to mitigate risk. One way to develop a hazard alerting system based on probabilistic models is by using a threshold-based approach, where the probability of the undesirable event without mitigation is compared against a...
READ MORE
U.S. Department of Transportation Federal Aviation Administration Field Demonstration #2: Final Report for Staffed NextGen Tower (SNT)
March 1, 2012
Project Report
Published in:
MIT Lincoln Laboratory Report ATC-389
Topic:
R&D area:
Summary
Staffed NextGen Towers (SNT), a research concept being developed and validated by the Federal Aviation Administration (FAA), is a paradigm shift to providing air traffic control services primarily via surface surveillance approved for operational use by controllers instead of the existing out-the-window (OTW) view at high-density airports. SNT was exercised as a prototype installed at the Dallas-Fortworth International Airport (DFW) during a two-week demonstration in the spring of 2011. MIT Lincoln Laboratory conducted this demonstration for the FAA in coordination with DFW air traffic control (ATC) and the DFW airport authority. This proof-of-concept demonstration used live traffic and was conducted by shadowing East tower operations from the DFW center tower, which is a back-up facility currently not typically used for air traffic control. The objective of this SNT field demonstration was to validate the supplemental SNT concept, to assess the operational suitability of the Tower Information Display System (TIDS) display for surface surveillance, and to evaluate the first iteration of prototype cameras in providing visual augmentation. TIDS provided surface surveillance information using an updated user interface that was integrated with electronic flight data. The cameras provided both fixed and scanning views of traffic to augment the OTW view. These objectives were met during the two-week field demonstration. DFW air traffic provided twelve controllers, three front line manager (FLMs), and three traffic management coordinators (TMCs) as test subjects. The twelve National Air Traffic Controllers Association (NATCA) DFW controllers "worked" the traffic according to their own techniques, using new hardware and software that included high resolution displays of surveillance data augmented by camera views. This equipment was designed to provide enhanced situational awareness to allow controllers to manage increased traffic volume during poor visibility conditions, leading to increased throughput. Results indicated that the likelihood of user acceptance and operational suitability is high for TIDS as a primary means for control, given surface surveillance that is approved for operational use. Human factors data indicated that TIDS could be beneficial. However, major technical issues included two display freezes, some incorrectly depicted targets, and display inconsistencies on TIDS. The cameras experienced numerous technical limitations that negatively influenced the human factors assessment of them. This report includes the percentages of human factors and technical success criteria that passed at DFW-2.
Summary
Staffed NextGen Towers (SNT), a research concept being developed and validated by the Federal Aviation Administration (FAA), is a paradigm shift to providing air traffic control services primarily via surface surveillance approved for operational use by controllers instead of the existing out-the-window (OTW) view at high-density airports. SNT was exercised...
READ MORE
Photonic ADC: overcoming the bottleneck of electronic jitter
February 13, 2012
Journal Article
Published in:
Opt. Express, Vol. 20, No. 4, 13 February 2012, pp. 4454-4469.
Topic:
R&D area:
Summary
Accurate conversion of wideband multi-GHz analog signals into the digital domain has long been a target of analog-to-digital converter (ADC) developers, driven by applications in radar systems, software radio, medical imaging, and communication systems. Aperture jitter has been a major bottleneck on the way towards higher speeds and better accuracy. Photonic ADCs, which perform sampling using ultra-stable optical pulse trains generated by mode-locked lasers, have been investigated for many years as a promising approach to overcome the jitter problem and bring ADC performance to new levels. This work demonstrates that the photonic approach can deliver on its promise by digitizing a 41 GHz signal with 7.0 effective bits using a photonic ADC built from discrete components. This accuracy corresponds to a timing jitter of 15 fs - a 4-5 times improvement over the performance of the best electronic ADCs which exist today. On the way towards an integrated photonic ADC, a silicon photonic chip with core photonic components was fabricated and used to digitize a 10 GHz signal with 3.5 effective bits. In these experiments, two wavelength channels were implemented, providing the overall sampling rate of 2.1 GSa/s. To show that photonic ADCs with larger channel counts are possible, a dual 20- channel silicon filter bank has been demonstrated.
Summary
Accurate conversion of wideband multi-GHz analog signals into the digital domain has long been a target of analog-to-digital converter (ADC) developers, driven by applications in radar systems, software radio, medical imaging, and communication systems. Aperture jitter has been a major bottleneck on the way towards higher speeds and better accuracy...
READ MORE
Diffractive beam combining of a 2.5-kW fiber laser array
January 29, 2012
Conference Paper
Published in:
ASSP 2012, Advanced Solid-State Photonics, 29 January - 1 February 2012.
Summary
Five 500-W fiber amplifiers were coherently combined with 79% efficiency using a diffractive optical element (DOE) combiner, generating a single beam whose M^2 = 1.1 beam quality exceeded that of the inputs.
Summary
Five 500-W fiber amplifiers were coherently combined with 79% efficiency using a diffractive optical element (DOE) combiner, generating a single beam whose M^2 = 1.1 beam quality exceeded that of the inputs.
READ MORE