The performances of the Viterbi and the interacting multiple model (IMM) algorithms applied to radar tracking and detection are investigated and compared. In the first case considered, target acceleration is identical to one of the system models, while in the second case it is not. The algorithm performance depends monotonically on the maximal magnitude of the difference between the model acceleration, on the time interval between measurements, and on the reciprocal of the measurement error standard deviation. In general, when these parameters are relatively large, both algorithms perform well. When they are relatively small, the Viterbi algorithm is better. However, during delay periods, namely right after the start of a maneuver, the IMM algorithm provides better estimations.>

The implementations of the Viterbi algorithm (VA) and the interacting multiple model (IMM) algorithm on a shared-bus and shared-memory multiple-input multiple-data (MIMD) multiprocessor are discussed. The computational complexity as well as the speedup and efficiency are examined in detail. It is shown that the computational complexity of the parallel implementation of these algorithms is about the same in both memory space and processing time categories. Efficiency with P processors is about 1-1/P for small P and is expected to be relatively high for large P, especially when many filters and large state and measurement vectors are considered.>

General high-resolution smoothing techniques for estimating deterministic sinusoidal frequencies from short-record noisy data are presented. These techniques are general in the sense that methods such as the modified least squares Prony method, as well as those which are based on eigenvector decompositions, may be considered as special cases of them. The theoretical basis of these smoothing techniques is discussed, and their performance in the presence of white Gaussian noise at low signal-to-noise ratio (SNR) is examined. It is shown that close to the threshold of the maximum-likelihood method (SNR approximately=3 dB) these symmetric smoothing techniques provide better accuracy than any other current method.>