https://hal.inria.fr/inria-00075657/document

Distance measures for signal processing and pattern recognition

Several properties, interrelations, and interpretations are developed for various speech spectral distortion measures. The principle results are 1) the development of notions of relative strength and equivalence of the various distortion measures both in a mathematical sense corresponding to subjective equivalence and in a coding sense when used in minimum distortion or nearest neighbor speech processing systems; 2) the demonstration that the Itakura-Saito and related distortion measures possess a property similar to the triangle inequality when used in nearest neighbor systems such as quantization and cluster analysis; and 3) that the Itakura-Saito and normalized model distortion measures yield efficient computation algorithms for generalized centroids or minimum distortion points of groups or clusters of speech frames, an important computation in both classical cluster analysis techniques and in algorithms for optimal quantizer design. We also argue that the Itakura-Saito and related distortions are well-suited computationally, mathematically, and intuitively for such applications.

Distortion measures for speech processing

Compression of digital electrocardiogram (ECG) signals is desirable for two reasons: economic use of storage space for data bases and reduction of the data transmission rate for compatibility with telephone lines. In a sample of 220 Frank4ead ECG's the removal of signal redundancy by second-order prediction or interpolation with subsequent entropy encoding of the respective residual errors was investigated. At the sampling rate of 200 Hz, interpolation provided a 6 dB smaller residual error variance than prediction. A near-optimal value for the interpolation coefficients is 0.5, permitting simple implementation of the algorithm and requiring a word length for arithmetic processing of only 2 bits in extent of the signal precision. For linear prediction, the effects of occasional transmission errors decay exponentially, whereas for interpolation they do not, necessitating error control in certain applications. Encoding of the interpolation errors by a Huffman code truncated to ±5 quantization levels of 30 ?V, required an average word length of 2.21 bits/sample (upper 96 percentile 3 bits/sample), resulting in data transmission rates of 1327 bits/s (1800 bits/s) for three simultaneous leads sampled at the rate of 200 Hz. Thus, compared with the original signal of 8 bit samples at 500 Hz, the average compression is 9:1. Encoding of the prediction errors required an average wordlength of 2.67 bits/sample with a 96 percentile of 5.5 bits/sample, making this method less suitable for synchronous transmission.

Compression of the ECG by Prediction or Interpolation and Entropy Encoding

Proactive network anomaly detection is crucial to provide quality of service guarantees for future networks. We characterize a set of proactively detectable anomalies in terms of the management information base variables. The time series data obtained from these variables are analyzed by an intelligent agent, which is a simple and lightweight signal processor. The agent provides real-time proactive alarms that are indicative of impending network problems and suggests the possibility of implementing automatic recovery mechanisms. Proactive detection was accomplished within minutes. The agent is shown to be amenable to distributed implementation, and is a promising approach to self-managed networks.

Proactive anomaly detection using distributed intelligent agents

The detection of network fault scenarios has been achieved using the statistical information contained in the Management Information Base (MIB) variables. An appropriate subset of MIB variables was chosen in order to adequately describe the function of the node. The time series data obtained from these variables was analyzed using a sequential generalized likelihood ratio (GLR) test. The GLR test was used to detect the change points in the behavior of the variables. Using a binary hypothesis test, variable level alarms were generated based on the magnitude of the detected changes as compared to the normal situation. These alarms were combined using a duration filter resulting in a set of node level alarms, which correlated with the experimentally observed network faults and performance problems. The algorithm has been tested on real network data. The applicability of our algorithm to a heterogeneous node was confirmed by using the MIB data from a second node. Interestingly, for most of the faults studied, detection occurs in advance of the fault (at least 5 min) and the algorithm is simple enough for potential online implementation: thus allowing the possibility of prediction and recovery in the future.

Adaptive thresholding for proactive network problem detection

This paper considers the problem of comparing two sets of (LPC) coefficients or, more generally, that of comparing two short segments of speech via LPC techniques. It is shown that Itakura's prediction-residual ratio is intuitively unsatisfactory and theoretically misleading as a distance measure. Two slower, but more accurate statistical means of comparison are suggested, and these are supported by evidence from a simulation study.

Statistical tests and distance measures for LPC coefficients

Partial inhibition of primary lung tumourigenesis by tetramisole.

This paper presents a stepwise method for analyzing the relationship between a set of discrete variables measured on a group of individuals, and an outcome such as disease incidence or prognosis. The method is based on information theory and is similar to some which have been used for investigating the order of Markov chains. A mixture of continuous and discrete variables can be studied and missing values estimated. An illustrative example is presented in which an assessment is made of the relationship between a set of factors and the incidence of heart attacks in patients on treatment for high blood pressure.

P. de Souza

Papers

Statistical tests and distance measures for LPC coefficients

Partial inhibition of primary lung tumourigenesis by tetramisole.

Stepwise information analysis of factors related to disease incidence and prognosis