- Research Article
1
- 10.1016/0165-1684(83)90073-7
Multilingual comparison of forward and backward adaptive prediction in DPCM
- May 01, 1983
- Signal Processing
- Louis C Sauter
Multilingual comparison of forward and backward adaptive prediction in DPCM
The formation of predictions is essential to our ability to build models of the world and use them for intelligent decision-making. Here we challenge the dominant assumption that humans form only forward predictions, which specify what future events are likely to follow a given present event. We demonstrate that in some environments, it is more efficient to use backward prediction, which specifies what present events are likely to precede a given future event. This is particularly the case in diverging environments, where possible future events outnumber possible present events. Correspondingly, in six preregistered experiments (n = 1,299) involving both simple decision-making and more challenging planning tasks, we find that humans engage in backward prediction in divergent environments and use forward prediction in convergent environments. We thus establish that humans adaptively deploy forward and backward prediction in the service of efficient decision-making.
Multilingual comparison of forward and backward adaptive prediction in DPCM
Multilingual comparison of forward and backward adaptive prediction in DPCM
Reciprocal Learning Networks for Human Trajectory Prediction
We observe that the human trajectory is not only forward predictable, but also backward predictable. Both forward and backward trajectories follow the same social norms and obey the same physical constraints with the only difference in their time directions. Based on this unique property, we develop a new approach, called reciprocal learning, for human trajectory prediction. Two networks, forward and backward prediction networks, are tightly coupled, satisfying the reciprocal constraint, which allows them to be jointly learned. Based on this constraint, we borrow the concept of adversarial attacks of deep neural networks, which iteratively modifies the input of the network to match the given or forced network output, and develop a new method for network prediction, called reciprocal attack for matched prediction. It further improves the prediction accuracy. Our experimental results on benchmark datasets demonstrate that our new method outperforms the state-of-the-art methods for human trajectory prediction.
Read moreExponential Modeling Using Combined Forward and Backward Prediction
: A sequence of complex data can be modeled using a damped exponential model. Forward or backward linear prediction can be used to estimate poles, and then total least squares can be used to estimate amplitudes for each pole. When model order is larger than true order, extraneous poles appear. These poles generally appear inside or outside the unit circle, depending on whether forward or backward prediction is used, respectively. Since true poles appear in approximately the same place for both methods, use of combined forward and backward pole estimation can be used to distinguish true poles from extraneous ones.
Read moreEmploying Volterra filters in the ADPCM technique for speech coding: a comprehensive investigation
Although linear filters are useful in a various applications in the context of speech processing, there are several evidences for existence of nonlinearity in speech signals. Our main aim is to launch a comprehensive investigation into the exploitation of nonlinear Volterra filters in the context of the ADPCM‐based speech coding technique, using two methods of forward prediction, based on the LS criterion, and backward prediction, based on both LMS and RLS adaptation algorithms. In any case, after solving some innate problems, for example, ill‐conditioning and instability, schemes for optimum exploitation of nonlinear prediction are developed and simulation results are provided, tested with several performance criteria. With forward prediction a scheme is developed to detect and flag those frames for which, after stabilizing, including the quadratic predictor is beneficial. Scalar and vector quantisation methods are used for quantising the residual signal and the filter parameters, respectively. The results show that using this scheme a negligible improvement (up to 0.62 dB in the SNR) can be achieved, in spite of the increase in bit rate and complexity. With backward prediction two frame‐based schemes are developed in which for each frame, after examining a set of quadratic filters, the best filter in the sense of the best quality of the reconstructed speech is selected. The ultimate schemes result in an improvement of up to 1.5 dB in the overall SNR of the reconstructed speech at the cost of a slight increase in the bit‐rate, a short delay and a demanding increase in the complexity. Copyright © 2010 John Wiley & Sons, Ltd.
Read moreSplit-dimension vector quantization of Parcor coefficients for low bit rate speech coding
A novel split vector quantization (SVQ) scheme for low bit rate coding of speech signals is proposed. In this scheme, the LPC parameter vector, which is represented by Parcor coefficients, is split into small-dimension subvectors, and each subvector is sequentially quantized according to a multistage structure that resembles a segmented lattice filter. The forward and backward prediction residuals in the segmented filter are coupled across VQ stages. The quantizer in each stage operates on the principle of minimizing the forward and backward prediction error energies similar to linear predictive analysis. Simulation results show that the new split VQ scheme can achieve transparent quantization of LPC parameters at 25 b/frame.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
Read moreRanging a blooming picture by using partial autocorrelation
A new method of designing control sensors for intelligent transport systems (ITS) is proposed in which blooming is estimated by partial autocorrelation analysis, the distance from the focal point to the object is derived, and the distance to the object is measured. In partial autocorrelation analysis, forward prediction and backward prediction are carried out simultaneously so that the point spread function convoluted with the blooming image is linearly predicted and the amount of blooming is estimated. The two-dimensional analysis requires a significant amount of processing and highly accurate operation. Therefore, in this method, by preprocessing in the frequency dimension, only the horizontal components are extracted for analysis. From the experimental results, it is found that distance estimation from a single image is possible near the focal length (2.5 m–1.7 m). © 1999 Scripta Technica, Electron Comm Jpn Pt 2, 82(1): 23–33, 1999
Read morePrediction of a glucose appearance function from foods using deconvolution
The glycaemic response of an insulin-treated diabetic patient goes through many transitory phases, leading to a steady state glycaemic profile following a change in either insulin regimen or diet. Most models attempting to model the glucose and insulin relationship try to model the effect of oral or injected glucose rather than that from the digestion of food. However, it is clear that a better understanding of the glycaemic response would arise from consideration of intestinal absorption from the gut. It is assumed that this type of absorption can be modelled by a so-called glucose appearance function (systemic appearance of glucose via glucose absorption from the gut) predicting the glucose load from the food. Much research has been carried out in the areas of hepatic balance, insulin absorption and insulin independent/dependent utilization. However, little is known about intestinal absorption patterns or their corresponding glucose appearance profiles. The strategy under investigation herein is to use deconvolution or backward engineering. By starting with specific results i.e. blood glucose and insulin therapy, it is possible to work backwards to predict the glucose forcing functions responsible for the outcome. Assuming compartmental consistency, this will allow a clearer insight into the true gut absorption process. If successful, the same strategy can be applied to more recent glucose and insulin models to further our understanding of the food to blood glucose problem. This paper investigates the Lehmann-Deutsch modified model of glucose and insulin interaction, created from the model proposed by Berger-Rodbard. The model attempts to simulate the steady state glycaemic and plasma insulin responses, independent of the initial values from which the simulation is started. Glucose enters the model via both intestinal absorption and hepatic glucose production. We considered a 70 kg male insulin-dependent diabetic patient with corresponding hepatic and insulin sensitivity parameters of 0.6 and 0.3 respectively. Net hepatic glucose balance was modelled piecewise by linear and symmetric functions. A first-order Euler method with step size of 15 minutes was employed. For the simulation, only Actrapid and NPH injections were considered. The injection of insulin and the glucose flux to the gut were started simultaneously to avoid any delay associated with gastric emptying. The systemic appearance of glucose was compared from two view points, not only to assess the strategic principle, but also to assess the suitability of the modifications made by Lehmann and Deutsch. The first is a forward prediction using the compartmental structure. This analysis involves the rate of gastric emptying without time delay. The second is a backward prediction from experimentally observed blood glucose profiles. Investigations involved porridge, white rice and banana containing the same carbohydrate content (25 g). Results obtained from the first analysis were dependent on the rate of gastric emptying, especially its ascending and descending branches. Results from the second analysis were dependent on the dose and type of insulin administered. Both predicted profiles showed consistency with physiological reasoning, although it became apparent that such solutions could be unstable. Furthermore, both types of prediction were similar in structure and appearance, especially in simulations for porridge and banana. This emphasized the consistency and suitability of both analyses when investigating the compartmental accuracy and limitations within a model. The new strategic approach was deemed a success within the model, and the modifications made by Lehmann and Deutsch appropriate. We suggest that a gastric emptying curve with a possible gastric delay is the way forward in regulating the appearance of glucose via gut absorption. The Lehmann-Deutsch gastric curve is described by either a trapezoidal or triangular function dependent on the carbohydrate cont
Read moreMedian type filters with linear predictive substructures
In this paper, we introduce a new class of FIR Median Hybrid (FMH) filters which contain linear FIR substructures to estimate the current signal value using forward and backward prediction. The output of the overall filter is the median of the predicted values and the actual signal value in the middle of the filter window. Predictors maximizing the signal to noise ratio on signal sections described by an 1th order polynomial are derived. The ramp enhancement filters are shown to attenuate the noise on a ramp signal better than the Standard Median (SM) filters. The new predictive FMH filters are shown to have root signals which do not exist for the SM filters, e.g. triangular waves. By combining the level and the ramp enhancement FMH filters, we obtain a filter which attenuates noise on constant and ramp signals.
Read moreA Comparison of Digital Video Standards
In 1992, MPEG-1 was created as the first standard for encoding moving pictures accompanied by sound. The aim was to achieve a picture quality close to that of VHS at CD data rates (< 1.5 Mbit/s). MPEG-1 was provided only for applications on storage media (CD, hard disk) and not for transmission (broadcasting) and its data structures correspond to this objective. The audio and video coding of MPEG-1 is quite close to that of MPEG-2 and all the fundamental algorithms and methods are already in place. There are both I, P and B frames, i.e. forward and backward prediction, and naturally there are the DCT-based irrelevance reduction methods already found in JPEG. The picture resolution, however, is limited to about half the VGA resolution (352 × 288). Neither is there any necessity for field encoding (interlaced scanning method). In MPEG-1, there is only the so-called Program Stream (PS) which is composed of multiplexed packetized elementary stream (PES) packets of audio and video. The variable-length (64 kbytes max) audio and video PES packets are simply assembled interleaved in accordance with the present data rate to form a data stream. This data stream is not processed any further since it is only intended to be stored on storage media and not used for transmission. A certain number of audio and video PES packets are combined to form a so-called pack which consists of a header and the payload just like the PES packets themselves. A pack is often based on the size of a physical data sector of the storage medium.
Read moreComplexity reduction method for High Efficiency Video Coding encoding based on scene-change detection and image texture information
Recently, as demand for high-quality video and realistic media has increased, High Efficiency Video Coding has been standardized. However, High Efficiency Video Coding requires heavy cost in terms of computational complexity to achieve high coding efficiency, which causes problems in fast coding processing and real-time processing. In particular, High Efficiency Video Coding inter-coding has heavy computational complexity, and the High Efficiency Video Coding inter prediction uses reference pictures to improve coding efficiency. The reference pictures are typically signaled in two independent lists according to the display order, to be used for forward and backward prediction. If an event occurs in the input video, such as a scene change, the inter prediction performs unnecessary computations. Therefore, the reference picture list should be reconfigured to improve the inter prediction performance and reduce computational complexity. To address this problem, this article proposes a method to reduce computational complexity for fast High Efficiency Video Coding encoding using information such as scene changes obtained from the input video through preprocessing. Furthermore, reference picture lists are reconstructed by sorting the reference pictures by similarity to the current coded picture using Angular Second Moment, Contrast, Entropy, and Correlation, which are image texture parameters from the input video. Simulations are used to show that both the encoding time and coding efficiency could be improved simultaneously by applying the proposed algorithms.
Read morePerformance of DF techniques with a VHF antenna array
A description of practical data processing techniques that have been used for direction finding (DF) with a real array is presented. The measured data were collected using a DF trial "test bed", which was developed at the Defence Research Establishment Ottawa (DREO) and is based on digital beamforming (DBF) technology. The system transmitters are used to simulate a complex signal environment and can be operated in a continuum of frequencies in the VHF band. In the paper, we present (1) a signal model, (2) multipath analysis, (3) array error calibration, (4) spatio-temporal estimation using multiple snapshots, (5) multiple snapshot forward backward two-step prediction DF, and (6) performance evaluation of DF algorithm. The DF results for a wideband radio source are also included.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">></ETX>
Read moreSelf-Constrained Inference Optimization on Structural Groups for Human Pose Estimation
We observe that human poses exhibit strong group-wise structural correlation and spatial coupling between keypoints due to the biological constraints of different body parts. This group-wise structural correlation can be explored to improve the accuracy and robustness of human pose estimation. In this work, we develop a self-constrained prediction-verification network to characterize and learn the structural correlation between keypoints during training. During the inference stage, the feedback information from the verification network allows us to perform further optimization of pose prediction, which significantly improves the performance of human pose estimation. Specifically, we partition the keypoints into groups according to the biological structure of human body. Within each group, the keypoints are further partitioned into two subsets, high-accuracy proximal keypoints and low-accuracy distal keypoints. We develop a self-constrained prediction-verification network to perform forward and backward predictions between these keypoint subsets. One fundamental challenge in pose estimation, as well as in generic prediction tasks, is that there is no mechanism for us to verify if the obtained pose estimation or prediction results are accurate or not, since the ground truth is not available. Once successfully learned, the verification network serves as an accuracy verification module for the forward pose prediction. During the inference stage, it can be used to guide the local optimization of the pose estimation results of low-accuracy keypoints with the self-constrained loss on high-accuracy keypoints as the objective function. Our extensive experimental results on benchmark MS COCO and CrowdPose datasets demonstrate that the proposed method can significantly improve the pose estimation results.KeywordsHuman pose estimationSelf-constrainedStructural inferencePrediction optimization
Read moreOn rigorously quantifying uncertainty in shear and compression wave velocity during surface wave site characterization
The accurate prediction of site effects is closely related to the quality of the seismic site characterization. Furthermore, increasingly probabilistic design procedures require not only a single seismic velocity profile, but a suite of profiles that rigorously accounts for epistemic and aleatory uncertainties. The quantification of these uncertainties is particularly important for non-invasive site characterization techniques that require inversion, such as surface wave methods, as the non-uniqueness of the derived velocity profiles depends on the uncertainties of the experimental data, which are dependent on the type and number of non-invasive measurements as well as the site's geologic setting. This work expands upon previous efforts to develop shear wave velocity (Vs) profile’s whose corresponding uncertainty is consistent with that of the experimental data and that rigorously address the non-uniqueness in the inverse problem. In particular, this work addresses three key points. First, how experimental data from multiple active-source and/or passive-wavefield surface wave arrays can be statistically combined to quantify site-specific aleatory uncertainty. Second, how multiple inversion parameterizations can and should be used to quantify epistemic uncertainty in the inverse problem. Third, how site-specific statistical models can be derived from non-invasive site characterization efforts as an alternative to existing Vs randomization techniques commonly used in practice. These advances are demonstrated at the Garner Valley Downhole Array (GVDA) site located in southern California. As a result of this study, new uncertainty-consistent shear and compression wave velocity profiles have been produced at the GVDA site for use in future forward and backward prediction efforts.
Read moreForward and backward prediction of process-tensile property of IN718 fabricated by selective laser melting
PurposeThe quality of Inconel 718 (IN718) from selective laser melting (SLM) is prerequisite for its application, and meeting required tensile properties is particularly important. This study aims to realize both mechanical property prediction and process parameter selection of SLM-ed IN718 by taking full advantage of their process-tensile property data mined from literature.Design/methodology/approachExtensive data of interest are mined from literature, among which the missing data are then imputed by fitting Gaussian mixture model via expectation maximization. Forward/backward predictive models for predicting the unknowns in tensile properties (ultimate tensile strength, yield strength, elongation) along horizontal and vertical directions and key process parameters (laser power, scanning speed, hatch spacing, layer thickness) are built through Bayesian network.FindingsNone of the experiments from literature has complete data of the four key process parameters and three tensile properties along two directions. Satisfactory accuracies are obtained for both data imputation for the missing values in the mined literature data with an average R2 of 0.64 and forward/backward prediction of process-tensile property with an average R2 of 0.58/0.54. The data imputation and predictive models are also tested with consistent prediction accuracies.Originality/valueForward/backward process-tensile property predictive models of SLM-ed IN718 with satisfactory performance can be obtained after data imputation for the mined literature data. Such models consider more process parameters (four key process parameters) and properties (three tensile properties along two directions), which also cover wider ranges than any individual studies through a less costly while effective approach.
Read moreModeling of Two Dimensional Fields with Autoregressive Lattice Parameters
The theory of one dimensional autoregressive lattice parameters, which has been successful in modeling signals such as speech (the voice tract), is extended to the modeling of two dimensional fields. Applications include lattice parameter modeling for purposes of identification, two dimensional digital filter design, and two dimensional spectral estimation. The concept of minimizing forward and backward prediction error sequences is extended to prediction error fields, and the lattice parameter formulation results which can be used for purposes of two dimensional data analysis, as well as for two dimensional signal synthesis and filtering.
Read more