Publications
Predicting the risk of complications in coronary artery bypass operations using neural networks
January 1, 1994
Conference Paper
Published in:
Proc. 7th Int. Conf. on Neural Information Processing Systems, NIPS, 1994, pp. 1055-62.
Topic:
R&D area:
R&D group:
Summary
Experiments demonstrated that sigmoid multilayer perceptron (MLP) networks provide slightly better risk prediction than conventional logistic regression when used to predict the risk of death, stroke, and renal failure on 1257 patients who underwent coronary artery bypass operations at the Lahey Clinic. MLP networks with no hidden layer and networks with one hidden layer were trained using stochastic gradient descent with early stopping. MLP networks and logistic regression used the same input features and were evaluated using bootstrap sampling with 50 replications. ROC areas for predicting mortality using preoperative input features were 70.5% for logistic regression and 76.0% for MLP networks. Regularization provided by early stopping was an important component of improved performance. A simplified approach to generating confidence intervals for MLP risk predictions using an auxiliary "confidence MLP" was developed. The confidence MLP is trained to reproduce confidence intervals that were generated during training using the outputs of 50 MLP networks trained with different bootstrap samples.
Summary
Experiments demonstrated that sigmoid multilayer perceptron (MLP) networks provide slightly better risk prediction than conventional logistic regression when used to predict the risk of death, stroke, and renal failure on 1257 patients who underwent coronary artery bypass operations at the Lahey Clinic. MLP networks with no hidden layer and networks...
READ MORE
Encoding approaches for data link transmission of weather graphics
December 10, 1993
Project Report
Published in:
MIT Lincoln Laboratory Report ATC-205
Summary
To provide pilots with necessary information to make informed decisions on the avoidance of hazardous weather and to maintain situational awareness of the weather conditions, the FAA is actively developing the capability to provide real-time graphical weather information to aircraft through the use of bandwidth-limited data links such as Mode S. The information content of weather images and the restricted bandwidth of the transmission channel require that the images be extensively compressed. This paper provides the results of a study concerning the applicability of various data compression algorithms to the weather image problem. Its conclusion is that the Polygon-Ellipse Algorithm developed at Lincoln Laboratory provides the best combination of compression, computational efficiency, and image quality for the encoding of weather images over the Mode S data link or other similarly bit-limited data links.
Summary
To provide pilots with necessary information to make informed decisions on the avoidance of hazardous weather and to maintain situational awareness of the weather conditions, the FAA is actively developing the capability to provide real-time graphical weather information to aircraft through the use of bandwidth-limited data links such as Mode...
READ MORE
Figure of merit training for detection and spotting
November 29, 1993
Conference Paper
Published in:
Proc. Neural Information Processing Systems, NIPS, 29 November - 2 December 1993.
Topic:
R&D area:
Summary
Spotting tasks require detection of target patterns from a background of richly varied non-target inputs. The performance measure of interest for these tasks, called the figure of merit (FOM), is the detection rate for target patterns when the false alarm rate is in an acceptable range. A new approach to training spotters is presented which computes the FOM gradient for each input pattern and then directly maximizes the FOM using back propagation. This eliminates the need for thresholds during training. It also uses network resources to model Bayesian a posteriori probability functions accurately only for patterns which have a significant effect on the detection accuracy over the false alarm rate of interest. FOM training increased detection accuracy by 5 percentage points for a hybrid radial basis function (RBF) - hidden Markov model (HMM) wordspotter on the credit-card speech corpus.
Summary
Spotting tasks require detection of target patterns from a background of richly varied non-target inputs. The performance measure of interest for these tasks, called the figure of merit (FOM), is the detection rate for target patterns when the false alarm rate is in an acceptable range. A new approach to...
READ MORE
Machine Intelligent Gust Front Algorithm
November 4, 1993
Project Report
Published in:
MIT Lincoln Laboratory Report ATC-196
Topic:
R&D area:
Summary
The Federal Aviation Administration has sponsored research and development of algorithms for automatic gust front detection as part of a suite of hazardous weather detection capabilities for airports. These algorithms are intended for use with Doppler radar systems, specifically the Terminal Doppler Weather Radar (TDWR) and the Airport Surveillance Radar enhanced with a Wind Shear Processor (ASR-9 WSP). Although gust fronts are observable with fairly reliable signatures in TDWR data, existing gust front detection algorithms have achieved only modest levels of detection performance. For smaller airports not slated to receive a dedicated TDWR, the ASR-9 WSP will provide a less expensive wind shear detection capability. Gust front detection in ASR-9 SP data is an even more difficult problem, given the reduced sensitivity and less reliable Doppler measurements of this radar. A Machine Intelligent Gust Front Algorithm (MIGFA) has been constructed at Lincoln Laboratory that is a radical departure from previous design strategies. Incorporating knowledge-based, signal-processing techniques initially developed at Lincoln Laboratory for automatic target recognition, MIGFA uses meterological knowledge, spatial and temporal context, conditional data fusion, delayed thresholding, and pixel-level fusion of evidence to improve gust front detection performance significantly. In tests comparing MIGFA with an existing state-of-the-art algorithm applied to ASR-9 WSP data, MIGFA has substantially outperformed the older algorithm. In fact, by some measures, MIGFA has done as well or better than human interpreters of the same data. Operational testing of this version was done during 1992 in Orlando, Florida. The desing, test results, and performance evaluation of hte ASR-9 WSP version of MIGFA are presented in this report, which was prepared as part of the documentation package for the ASR-9 WSP gust front algorithm.
Summary
The Federal Aviation Administration has sponsored research and development of algorithms for automatic gust front detection as part of a suite of hazardous weather detection capabilities for airports. These algorithms are intended for use with Doppler radar systems, specifically the Terminal Doppler Weather Radar (TDWR) and the Airport Surveillance Radar...
READ MORE
ADS-Mode S system overview
October 25, 1993
Conference Paper
Published in:
AIAA/IEEE Digital Avionics Systems Conf., 25-28 October 1993, pp. 104-109.
Topic:
R&D area:
R&D group:
Summary
ADS-Mode S is a system concept that merges the capabilities of Automatic Dependent Surveillance and the Mode S beacon radar. The result is an integrated system for seamless surveillance and data link that permits equipped aircraft to participate in ADS or beacon ground environments. This offers many possibilities for transitioning from a beacon to an ADS based surveillance system. The ADS-Mode S squitter. The current Mode S squitter is a spontaneous, periodic (once per second) 56-bit broadcast message containing the Mode S 24-bit address. This broadcast is provided by all Mode S transponders and is used by the Traffic Alert and Collision Avoidance System (TCAS) to acquire nearby Mode S equipped aircraft. For ADS-Mode S use, this squitter broadcast would be extended to 112 bits to provide for the transmission of a 56-bit ADS message field. The ADS squitter would be transmitted in addition to the current TCAS squitter in order to maintain compatibility with current TCAS equipment during transition.
Summary
ADS-Mode S is a system concept that merges the capabilities of Automatic Dependent Surveillance and the Mode S beacon radar. The result is an integrated system for seamless surveillance and data link that permits equipped aircraft to participate in ADS or beacon ground environments. This offers many possibilities for transitioning...
READ MORE
ADS-Mode S
October 24, 1993
Conference Paper
Published in:
Proc. 38th Annual Air Traffic Control Association Convention, 24-28 October 1993, pp. 230-236.
Topic:
R&D area:
R&D group:
Summary
The International Civil Aviation Organization (ICAO) has defined a concept for communications, navigation, and surveillance for the next century known as the Future Air Navigation System (FANS). A cornerstone, of the FANS is an increasing reliance on satellite-based position-determining systems such as the Global Positioning System (GPS). In the case of, surveillance, aircraft position information is automatically downlinked to ground controllers. This technique is known as Automatic Dependent Surveillance (ADS). ADS-Mode S is an ADS system concept utilizing the frequencies and formats of the Mode S system for downlinking position information and also uplinking differential GPS (DGPS) corrections. The result is an integrated concept for surveillance that permits aircraft equipped with a Mode S transponder and a GPS receiver to participate in both ADS and beacon ground environments. This makes possible a smooth transition of the National Airspace System (NAS) secondary surveillance system from a beacon-based to an ADS-based environment. In addition, several other benefits from ADS-Mode S accrue to the Traffic Alert and Collision Avoidance System (TCAS) and to the Mode S Data Link system.
Summary
The International Civil Aviation Organization (ICAO) has defined a concept for communications, navigation, and surveillance for the next century known as the Future Air Navigation System (FANS). A cornerstone, of the FANS is an increasing reliance on satellite-based position-determining systems such as the Global Positioning System (GPS). In the case...
READ MORE
ASR-9 Microburst Detection Algorithm
October 22, 1993
Project Report
Published in:
MIT Lincoln Laboratory Report ATC-197
Summary
The ASR-9 Wind Shear Processor (WSP) is intended as an economical alternative for those airports that have not been slated to receive a Terminal Doppler Weather Radar (TDWR) but have, or will be receiving, an ASR-9 radar. Lincoln Laboratory has developed a prototype ASR-9 WSP system which has been demonstrated during the summer months of the past three year in Orlando, Florida. During the operational test period, microburst and gust front warnings, as well as storm motion indications, were provided to the Air Traffic Control in real time. The ASR-9 Microburst Detection Algorithm (AMDA) is based on the earlier TDWR Microburst Detection Algorithm but has been substantially modified to match better the particular strengths and weaknesses of the ASR-9 rapid-scanning fan-beam radar. The most significant additions included a capability to detect overhead microbursts, a reflectivity processing step used to help detect velocity signatures that have been biased by overhanging precipitation, and a modification to some of the shear segment grouping and thresholding parameters to accommodate better the typical on-air siting of the ASR-9. In addition, the AMDA has been designed to be as efficient as possible to allow it to run at the radar's 4.8 seconds/scan antennas rotation rate on a single-board computer. A detailed description of AMDA, as well as the performance evaluation strategy and results, are presented in this report.
Summary
The ASR-9 Wind Shear Processor (WSP) is intended as an economical alternative for those airports that have not been slated to receive a Terminal Doppler Weather Radar (TDWR) but have, or will be receiving, an ASR-9 radar. Lincoln Laboratory has developed a prototype ASR-9 WSP system which has been demonstrated...
READ MORE
CW operation of monolithic arrays of surface-emitting AlGaAs diode lasers with dry-etched vertical facets and parabolic deflecting mirrors
October 1, 1993
Journal Article
Published in:
IEEE Photonics Technol. Lett., Vol. 5, No. 10, October 1993, pp. 1146-1149.
Topic:
R&D area:
Summary
A monolithic two-dimensional array of surface-emitting AlGaAs diode lasers with dry-etched vertical facets and parabolic deflecting mirrors was mounted junction-side up on a W / Cu microchannel heatsink and evaluated under continuous-wave (CW) operating conditions. Both the facets and parabolic deflecting mirrors were etched using chlorine ion-beam-assistd etching. Threshold current densities of different sections of the array were consistently around 240 A/cm (to the second power), and measured CW differential quantum efficiencies were in the 46-48% range. CW power densities as high as 148 W/cm (to the second power) were achieved with an average temperature rise of less than 25 degrees C in this junction-side-up configuration.
Summary
A monolithic two-dimensional array of surface-emitting AlGaAs diode lasers with dry-etched vertical facets and parabolic deflecting mirrors was mounted junction-side up on a W / Cu microchannel heatsink and evaluated under continuous-wave (CW) operating conditions. Both the facets and parabolic deflecting mirrors were etched using chlorine ion-beam-assistd etching. Threshold current...
READ MORE
The gust-front detection and wind-shift algorithms for the Terminal Doppler weather radar system
October 1, 1993
Journal Article
Published in:
J. Atmos. Ocean. Technol., Vol. 10, October 1993, pp. 693-709.
Topic:
R&D area:
R&D group:
Summary
The Federal Aviation Administration's Terminal Doppler Weather Radar (TDWR) system was primarily designed to address the operational needs of pilots in the avoidance of low-altitude wind shears upon takeoff and landing at airports. One of the primary methods of wind-shear detection for the TDWR system is the gust-front detection algorithm. The algorithm is designed to detect gust fronts that produce a wind-shear hazard and/or sustained wind shifts. It serves the hazard warning function by providing an estimate of the wind-speed gain for aircraft penetrating the gust front. The gust-front detection and wind-shift algorithms together serve a planning function by providing forecasted gust-front locations and estimates of the horizontal wind vector behind the front, respectively. This information is used by air traffic managers to determine arrival and departure runway configurations and aircraft movements to minimize the impact of wind shifts on airport capacity. This paper describes the gust-front detection and wind-shift algorithms to be fielded in the initial TDWR systems. Results of a quantitative performance evaluation using Doppler radar data collected during TDWR operational demonstrations at the Denver, Kansas City, and Orlando airports are presented. The algorithms were found to be operationally useful by the FAA airport controllers and supervisors.
Summary
The Federal Aviation Administration's Terminal Doppler Weather Radar (TDWR) system was primarily designed to address the operational needs of pilots in the avoidance of low-altitude wind shears upon takeoff and landing at airports. One of the primary methods of wind-shear detection for the TDWR system is the gust-front detection algorithm...
READ MORE
Energy separation in signal modulations with application to speech analysis
October 1, 1993
Journal Article
Published in:
IEEE Trans. Signal Process., Vol. 41, No. 10, October 1993, pp. 3024-3051.
Topic:
R&D area:
Summary
Oscillatory signals that have both an amplitude-modulation (AM) and a frequency-modulation (FM) structure are encountered in almost all communication systems. We have also used these structures recently for modeling speech resonances, being motivated by previous work on investigating fluid dynamics phenomena during speech production that provide evidence for the existence of modulations in speech signals. In this paper, we use a nonlinear differential operator that can detect modulations in AM-FM signals by estimating the product of their time-varying amplitude and frequency. This operator essentially tracks the energy needed by a source to produce the oscillatory signal. To solve the fundamental problem of estimating both the amplitude envelope and instantaneous frequency of an AM-FM signal we develop a novel approach that uses nonlinear combinations of instantaneous signal outputs from the energy operator to separate its output energy product into its amplitude modulation and frequency modulation components. The theoretical analysis is done first for continuous-time signals. Then several efficient algorithms are developed and compared for estimating the amplitude envelope and instantaneous frequency of discrete-time AM-FM signals. These energy separation algorithms are then applied to search for modulations in speech resonances, which we model using AM-FM signals to account for time-varying amplitude envelopes and instantaneous frequencies. Our experimental results provide evidence that bandpass filtered speech signals around speech formants contain amplitude and frequency modulations within a pitch period. Overall, the energy separation algorithms, due to their very low computational complexity and instantaneously-adapting nature, are very useful in detecting modulation patterns in speech and other time-varying signals.
Summary
Oscillatory signals that have both an amplitude-modulation (AM) and a frequency-modulation (FM) structure are encountered in almost all communication systems. We have also used these structures recently for modeling speech resonances, being motivated by previous work on investigating fluid dynamics phenomena during speech production that provide evidence for the existence...
READ MORE