No description available.

International Telemetering Conference Proceedings / October 25-28, 1993 / Riviera Hotel and Convention Center, Las Vegas, Nevada / This paper describes algorithms implemented by the Magellan High Rate Processor to recover radar data corrupted by the failure of an onboard tape recorder that dropped bits. For data with error correction coding, an algorithm was developed that decodes data in the presence of bit errors and missing bits. For the SAR data, the algorithm takes advantage of properties in SAR data to locate corrupted bits and reduce there effects on downstream processing. The algorithms rely on communication approaches, including an efficient tree search and the Viterbi algorithm to maintain the required throughput rate.

We address the problem of multi-view multi-target tracking using multiple stationary cameras in the application of hockey tracking and test the approach with data from two cameras. The system is based on the previous work by Okuma et al. [50]. We replace AdaBoost detection with blob detection in both image coordinate systems after background subtraction. The sets of blob-detection results are then mapped to the rink coordinate system using a homography transformation. These observations are further merged into the final detection result which will be incorporated into the particle filter. In addition, we extend the particle filter to use multiple observation models, each corresponding to a view. An observation likelihood and a reference color model are also maintained for each player in each view and are updated only when the player is not occluded in that view. As a result of the expanded coverage range and multiple perspectives in the multi-view tracking, even when the target is occluded in one view, it still can be tracked as long as it is visible from another view. The multi-view tracking data are further processed by trajectory smoothing using the Maximum a posteriori smoother. Finally, automatic camera selection is performed using the Hidden Markov Model to create personalized video programs.

We address the problem of multi-view multi-target tracking using multiple stationary cameras in the application of hockey tracking and test the approach with data from two cameras. The system is based on the previous work by Okuma et al. [50]. We replace AdaBoost detection with blob detection in both image coordinate systems after background subtraction. The sets of blob-detection results are then mapped to the rink coordinate system using a homography transformation. These observations are further merged into the final detection result which will be incorporated into the particle filter. In addition, we extend the particle filter to use multiple observation models, each corresponding to a view. An observation likelihood and a reference color model are also maintained for each player in each view and are updated only when the player is not occluded in that view. As a result of the expanded coverage range and multiple perspectives in the multi-view tracking, even when the target is occluded in one view, it still can be tracked as long as it is visible from another view. The multi-view tracking data are further processed by trajectory smoothing using the Maximum a posteriori smoother. Finally, automatic camera selection is performed using the Hidden Markov Model to create personalized video programs.

We address the problem of multi-view multi-target tracking using multiple stationary cameras in the application of hockey tracking and test the approach with data from two cameras. The system is based on the previous work by Okuma et al. [50]. We replace AdaBoost detection with blob detection in both image coordinate systems after background subtraction. The sets of blob-detection results are then mapped to the rink coordinate system using a homography transformation. These observations are further merged into the final detection result which will be incorporated into the particle filter. In addition, we extend the particle filter to use multiple observation models, each corresponding to a view. An observation likelihood and a reference color model are also maintained for each player in each view and are updated only when the player is not occluded in that view. As a result of the expanded coverage range and multiple perspectives in the multi-view tracking, even when the target is occluded in one view, it still can be tracked as long as it is visible from another view. The multi-view tracking data are further processed by trajectory smoothing using the Maximum a posteriori smoother. Finally, automatic camera selection is performed using the Hidden Markov Model to create personalized video programs. / Science, Faculty of / Computer Science, Department of / Graduate

In this work, a new method of approximating the Maximum-likelihood estimate has been presented. The method consists of first using the Viterbi algorithm to estimate the log likelihood of the state, and then approximating that log likelihood to keep the computational complexity down. Various methods for approximating the log likelihood are introduced, most of these using linear regression and feature vectors. The methods were compared to a Kalman filter or Extended Kalman filter (depending on wether the system was linear or nonlinear) as well as a Particle filter modified to return a maximum likelihood estimate. Two systems were used for testing, one very simple linear system as well as a complex nonlinear system. Both of these were 1-dimensional. When applied to the simple system, the presented method outperformed both the Kalman filter and the Particle filter. While many approximation methods gave a good results the best one was using a cubic spline. For the more complex system, the method presented here could not outperform the particle filter. The most promising approximation method for this system was a Chebyshev approximation.

No description available.

International Telemetering Conference Proceedings / October 25-28, 1999 / Riviera Hotel and Convention Center, Las Vegas, Nevada / Gaussian Minimum Shift Keying (GMSK) is a form of Continuous Phase Modulation (CPM) whose spectral occupancy can be easily tailored to the available channel bandwidth by a suitable choice of signal parameters. The constant envelope of the GMSK signal enables it to corporate with saturated power amplifier without the spectral re-growth problem. This paper provides a quantitative synopsis of binary GMSK signals in terms of their bandwidth occupancy and coherent demodulation performance. A detailed account of how to demodulate such signals using the Viterbi Algorithm (VA) is given, along with analytical power spectral density (PSD) and computer simulated bit-error-rate (BER) results for various signal BT products. The effect of adjacent channel interference (ACI) is also quantified. Ideal synchronization for both symbol time and carrier phase is assumed.

Optimal decision directed equalization techniques for time dispersive communication channels are often too complex to implement. This thesis considers reduced complexity decision directed equalization that lowers complexity demands yet retains close to optimal performance. The first part of this dissertation consists of three reduced complexity algorithms based on the Viterbi Algorithm (VA) which are: the Parallel Trellis VA (PTVA); Time Reverse Reduced State Sequence Estimation (TR-RSSE); and Forward-Backward State Sequence Detection (FBSSD). The second part of the thesis considers structural modifications of the Decision Feedback Equalizer (DFE), which is a special derivative of the VA, specifically, optimal vector quantization for fractionally spaced DFEs, and extended stability regions for baud spaced DFEs using passivity analysis are investigated.¶ For a special class of sparse channels the VA can be decomposed over a number of independent parallel trellises. This decomposition will be called the Parallel Trellis Viterbi Algorithm and can have lower complexity than the VA yet it retains optimal performance. By relaxing strict sparseness constraints on the channel a sub-optimal approach is proposed which keeps complexity low and obtains good performance.¶ Reduced State Sequence Estimation (RSSE) is a popular technique to reduce complexity. However, its deficiency can be the inability to adequately equalize non-minimum phase channels. For channels that have energy peaks in the tail of the impulse response (post-cursor dominant) RSSE's complexity must be close to the VA or performance will be poor. Using a property of the VA which makes it invariant to channel reversal, TR-RSSE is proposed to extend application of RSSE to post-cursor dominant channels.¶ To further extend the class of channels suitable for RSSE type processing, FBSSD is suggested. This uses a two pass processing method, and is suited to channels that have low energy pre and post-cursor. The first pass generates preliminary estimates used in the second pass to aid the decision process. FBSSD can range from RSSE to TR-RSSE depending on parameter settings.¶ The DFE is obtained when the complexity of RSSE is minimized. Two characterizing properties of the DFE, which are addressed in this thesis, are feedback and quantization. A novel fractionally spaced (FS) DFE structure is presented which allows the quantizer to be generalized relative to the quantizer used in conventional FS-DFEs. The quantizer can be designed according to a maximum a posteriori criterion which takes into account a priori statistical knowledge of error occurrences. A radically different quantizer can be obtained using this technique which can result in significant performance improvements.¶ Due to the feedback nature of the DFE a form of stability can be considered. After a decision error occurs, a stable DFE will, after some finite time and in the absence of noise, operate error free. Passivity analysis provides sufficient conditions to determine a class of channels which insures a DFE will be stable. Under conditions of short channels and small modulation alphabets, it is proposed that conventional passivity analysis can be extended to account for varying operator gains, leading to weaker sufficient conditions for stability (larger class of channels).

Trellis-coded Modulation (TCM) is used in bandlimited communication systems. TCM efficiency improves coding gain by combining modulation and forward error correction coding in one process. In TCM, the bandwidth expansion is not required because it uses the same symbol rate and power spectrum; the differences are the introduction of a redundancy bit and the use of a constellation with double points. <p> In this thesis, a novel TCM encoder/decoder ASIC chip implementation is presented. This ASIC codec not only increases decoding speed but also reduces hardware complexity. The algorithm and technique are presented for a 16-state convolutional code which is used in standard 256-QAM wireless systems. In the decoder, a Hamming distance is used as a cost function to determine output in the maximum likelihood Viterbi decoder. Using the relationship between the delay states and the path state in the Trellis tree of the code, a pre-calculated Hamming distances are stored in a look-up table. In addition, an output look-up-table is generated to determine the decoder output. This table is established by the two relative delay states in the code. The thesis provides details of the algorithm and the structure of TCM codec chip. Besides using parallel processing, the ASIC implementation also uses pipelining to further increase decoding speed. <p> The codec was implemented in ASIC using standard 0.18Ým CMOS technology; the ASIC core occupied a silicon area of 1.1mm2. All register transfer level code of the codec was simulated and synthesized. The chip layout was generated and the final chip was fabricated by Taiwan Semiconductor Manufacturing Company through the Canadian Microelectronics Corporation. The functional testing of the fabricated codec was performed partially successful; the timing testing has not been fully accomplished because the chip was not always stable.