R. Haddad and T. Parsons. Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....(t 1) 2 value. Median based predictors are particularly useful if the measurements contain randomly occurring asymmetric outliers that are rejected. However, they lack some of the smoothing that occurs with a mean based method, possibly resulting in forecasts with a considerable amount of jitter [17]. A third class of common predictors is Autoregressive Models [34, 14, 17] These have been used by other predictive techniques, such as NWS. We used both mean based and average based predictors for our predictions. 3.2 Context Insensitive Factors More recent values are often better predictors ....

....the measurements contain randomly occurring asymmetric outliers that are rejected. However, they lack some of the smoothing that occurs with a mean based method, possibly resulting in forecasts with a considerable amount of jitter [17] A third class of common predictors is Autoregressive Models [34, 14, 17]. These have been used by other predictive techniques, such as NWS. We used both mean based and average based predictors for our predictions. 3.2 Context Insensitive Factors More recent values are often better predictors of future behavior than an entire data set, no matter which mathematical ....

R. Haddad and T. Parsons, Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.

....3. KLT and Dimension Reduction Signal transforms are frequently used in the field of digital signal processing. Transformed signals can yield better insight into certain properties. Well known orthonormal transforms are the Discrete Cosine Transform (DCT) and the Discrete Fourier Transform (DFT) [7]. Both of these transforms have a fixed set of orthonormal basis vectors, which means, that the basis vectors are independent of the actual signal to be processed. In contrast signal dependent transforms, such as the KarhunenLo eve Transform (KLT) are known. It s basis vectors are matched to the ....

....to the statistics of the input signal to deliver a set of uncorrelated coefficients in the transform domain. The KLT is the optimum least squares decorrelating transform and has been successfully applied to problems in the fields of speech processing and pattern recognition [8] It can be shown [7], that it concentrates a maximum amount of signal energy into a small number of orthogonal Eigenspaces. But when only a few signal dimensions contain a significant amount of signal energy, these dimensions are sufficient for an approximate reconstruction of the signal. Additive uncorrelated noise, ....

Richard A. Haddad and Thomas W. Parsons, Digital Signal Processing - Theory, Applications and Hardware, Computer Science Press, 1991.

....DFT technique and compare it with traditional scan based DFT techniques. VIII. EXPERIMENTAL RESULTS We applied the nonscan DFT algorithms presented in this paper to different types of data path circuits, synthesized using the high level synthesis system HYPER [39] from behavioral descriptions [48]. In this section, we report the results obtained on the following data paths: 1) fourth order IIR cascade filter (4IIRcas) 2) speech filter (Speech) 3) fifth order EWF; 4) fifth order EWF, synthesized using high hardware sharing (EWFhigh) 5) fourth order IIR parallel filter, synthesized ....

R. Haddad and T. Parsons, Digital Signal Processing: Theory, Applications and Hardware. New York: Computer Science, 1991.

....(t 1) 2 value. Median based predictors are particularly useful if the measurements contain randomly occurring asymmetric outliers that are rejected. However, they lack some of the smoothing that occurs with a mean based method, possibly resulting in forecasts with a considerable amount of jitter [20]. A third class of common predictors is auto regressive models [17, 20, 42] We use an Auto regressive Integrated Moving Average (ARIMA) model technique that is constructed using the equation: Y t = a bY t 1 , where Y t is the prediction for time, t, Y t 1 is the previous data occurrence and ....

....the measurements contain randomly occurring asymmetric outliers that are rejected. However, they lack some of the smoothing that occurs with a mean based method, possibly resulting in forecasts with a considerable amount of jitter [20] A third class of common predictors is auto regressive models [17, 20, 42]. We use an Auto regressive Integrated Moving Average (ARIMA) model technique that is constructed using the equation: Y t = a bY t 1 , where Y t is the prediction for time, t, Y t 1 is the previous data occurrence and a and b are the regression coefficients that are computed based on past ....

R. Haddad and T. Parsons, Digital Signal Processing: Theory, Applications, and Hardware, Computer Science Press, 1991.

....is preserved. V. EXPERIMENTAL RESULTS We applied the proposed controller based DFT technique, consisting of the two phases described in Sections III and IV, on several controller data path circuits, synthesized using the behavioral test synthesis system BETS [9] from behavioral descriptions [12]. In this section, we report the results obtained for the following circuits implementing: 1) a 4th order IIR filter with a word size of 8 bits (4IIR.8) 2) a 4th order IIR filter with a word size of 14 bits (4IIR.14) 2) a Speech filter of word size 12 bits (Speech.12) and (5) a Wave Digital ....

R.A. Haddad and T.W. Parsons. Digital Signal Processing: Theory, Applications and Hardware. Computer Science Press, New York, NY, 1991.

....of different domains using the measurements that have been taken between the single pair of hosts in the cliques that span domains. This hierarchical t 1 A B Figure 2. The TCP experiment organization permits the NWS intrusiveness control mechanisms to scale. The NWS uses time series analysis, [8, 1, 10] to predict the performance of the network. Each monitored host is expected periodically to conduct several network probes that give a measure of the latency and the throughput of the link between it and all other hosts. The more periodical the series is, the more accurate and statistically valid ....

R. Haddad and T. Parsons. Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.

....with a 64KB transfer. 2.1.2 JavaNws Forecasting Module To make predictions of near future performance we implemented the NWS forecasters in Java. A detailed description of the these forecasters can be found in [19] In short, this module consists of a set of independent forecasting algorithms [7, 1, 8], each of which produces a one step ahead forecast from a given time series. At each time step, the measurement data taken by the applet is compared to the forecast produced by each forecasting algorithm for that time step. The difference between each forecast and the measurement it is forecasting ....

R. Haddad and T. Parsons. Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.

....model has been shown to yield forecasts that are equivalent to, or slightly better than, the best forecaster in the set [29] To be ecient, each of the techniques must be relatively cheap to compute. We have borrowed heavily from methodologies used by the digital signal processing community [19] in our implementation. A complete description of each method and its relative advantages is provided in [29] 19] and 9 [16] Brie y summarized, each method uses a sliding window over previous measurements to compute a one step ahead forecast based either on some estimate of the mean or ....

....in the set [29] To be ecient, each of the techniques must be relatively cheap to compute. We have borrowed heavily from methodologies used by the digital signal processing community [19] in our implementation. A complete description of each method and its relative advantages is provided in [29] [19], and 9 [16] Brie y summarized, each method uses a sliding window over previous measurements to compute a one step ahead forecast based either on some estimate of the mean or median of those measurements. To evaluate the accuracy of each forecast, we examine two forms of error. The rst, ....

R. Haddad and T. Parsons. Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.

....model has been shown to yield forecasts that are equivalent to, or slightly better than, the best forecaster in the set [29] To be efficient, each of the techniques must be relatively cheap to compute. We have borrowed heavily from methodologies used by the digital signal processing community [19] in our implementation. A complete description of each method and its relative advantages is provided in [29] 19] and [16] Briefly summarized, each method uses a sliding window over previous measurements to compute a one stepahead forecast based either on some estimate of the mean or median ....

....the set [29] To be efficient, each of the techniques must be relatively cheap to compute. We have borrowed heavily from methodologies used by the digital signal processing community [19] in our implementation. A complete description of each method and its relative advantages is provided in [29] [19], and [16] Briefly summarized, each method uses a sliding window over previous measurements to compute a one stepahead forecast based either on some estimate of the mean or median of those measurements. To evaluate the accuracy of each forecast, we examine two forms of error. The first, given ....

R. Haddad and T. Parsons. Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.

....method for the resources we currently monitor with the NWS. 4.2 Median based Methods The median value can also serve as a useful predictor, particularly if the measurement sequence contains randomly occurring, asymmetric outliers. Our presentation of these techniques follows the exposition in [19] and [12] The median over a sliding window of fixed length whose leading edge is the most recent measurement is used as the forecast for the next measurement. That is, we define Sort K = the sorted sequence of the K most recent measurement values, Sort K (j) the jth value in the sorted ....

....Equation 9. Median filters are attractive because they will reject the effects of sharply outlying data points or impulses from the forecasts they produce. They lack some of the smoothing power of the averaging based methods, however, resulting in forecasts with a considerable amount of jitter [19]. It is possible to combine the positive advantages of both classes of methods in the form of an ff trimmed mean filter that averages the central K Gamma 2 ff K values within a sliding window of size K for (0 ff 0:5) We define T = bff Kc for window size K, and the trimmed mean to be ....

[Article contains additional citation context not shown here]

Haddad, R., and Parsons, T. Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.

....model has been shown to yield forecasts that are equivalent to, or slightly better than, the best forecaster in the set [29] To be efficient, each of the techniques must be relatively cheap to compute. We have borrowed heavily from methodologies used by the digital signal processing community [19] in our implementation. A complete description of each method and its relative advantages is provided in [29] 19] and [16] Briefly summarized, each method uses a sliding window over previous measurements to compute a one step ahead forecast based either on some estimate of the mean or ....

....the set [29] To be efficient, each of the techniques must be relatively cheap to compute. We have borrowed heavily from methodologies used by the digital signal processing community [19] in our implementation. A complete description of each method and its relative advantages is provided in [29] [19], and [16] Briefly summarized, each method uses a sliding window over previous measurements to compute a one step ahead forecast based either on some estimate of the mean or median of those measurements. To evaluate the accuracy of each forecast, we examine two forms of error. The first, ....

R. Haddad and T. Parsons. Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.

....and compare it with traditional scan based DFT techniques. 8 Experimental Results We applied the non scan design for testability algorithms presented in this paper to different types of data path circuits, synthesized using the high level synthesis system HYPER [44] from behavioral descriptions [24]. In this section, we report the results obtained on the following data paths: 1) 4th order IIR cascade filter (4IIRcas) 2) Speech filter (Speech) 3) 5th order elliptical wave digital filter (EWF) 4) 5th order elliptical wave digital filter, synthesized using high hardware sharing ....

R.A. Haddad and T.W. Parsons. Digital Signal Processing: Theory, Applications and Hardware. Computer Science Press, New York, NY, 1991.

....method for the resources we currently monitor with the NWS. 4.2. Median based Methods The median value can also serve as a useful predictor, particularly if the measurement sequence contains randomlyoccurring, asymmetric outliers. Our presentation of these techniques follows the exposition in [19] and [12] The median over a sliding window of fixed length whose leading edge is the most recent measurement is used as the forecast for the next measurement. That is, we define SortK = the sorted sequence of the K most recent measurement values, SortK (j) the jth value in the sorted ....

....Equation 9. Median filters are attractive because they will reject the effects of sharply outlying data points or impulses from the forecasts they produce. They lack some of the smoothing power of the averaging based methods, however, resulting in forecasts with a considerable amount of jitter [19]. It is possible to combine the positive advantages of both classes of methods in the form of an ff trimmed mean filter that averages the central K Gamma 2 ff K values within a sliding window of size K for (0 ff 0:5) We define T = bff Kc for window size K , and the trimmed mean to be ....

[Article contains additional citation context not shown here]

R. Haddad and T. Parsons. Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.

No context found.

R. Haddad and T. Parsons. Digital Signal Processing: Theory, Applications, and Hardware. Computer Science Press, 1991.

No context found.

R.A. Haddad and T.W. Parsons. Digital Signal Processing: Theory, Applications, and Hardware. New York: Computer Science Press, 1991.

No context found.

R.A. Haddad and T.W. Parsons. Digital Signal Processing: Theory, Applications and Hardware. Computer Science Press, New York, NY, 1991.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC