Biomedical Signal Processing

Volume 2: Compression and Automatic Recognition

Arnon Cohen

• Publisher: CRC Press
• Year: 2019
• ISBN:
  9781000697506
• Edition: 1
• Topic: Scienze biologiche
• Subtopic: Microbiologia

First published in 1986: The presentation of the material in the book follows the flow of events of the general signal processing system. After the signal has been acquired, some manipulations are applied in order to enhance the relevant information present in the signal. Simple, Optimal, and adaptive filtering are examples of such manipulations. The detection of wavelets is of importance in biomedical signals; they can be detected from the enhanced signal by several methods. The signal very often contains redundancies. When effective storing, transmission, or automatic classification are required, these redundancies have to be extracted.

Chapter 1

WAVELET DETECTION

I. INTRODUCTION

A common problem in biomedical signal processing is that of detecting the presence of a wavelet in a noisy signal. A wavelet is considered here as any real valued function of time, possessing some structure. The approximate shape of the wavelet, expected to be present in the noisy signal, may be known and the requirements are to estimate its time of occurrence and its exact shape. The problem, of course, is not restricted to biomedical signals; it appears, for example, in communications, radar, sonar, or seismological signals. Two main approaches are used to solve the problem. The first uses structural features of the wavelet and the second uses template matching techniques.

The approximate knowledge of the wavelet’s shape can be used to select typical structural features. These features are random variables with some empirically determined probability density functions. The noisy signal is then continuously searched for these features.

The search can be conducted in various ways. Sophisticated algorithms that can be used to recognize a wavelet, once its features have been extracted, are discussed in Chapters 3 and 4. These algorithms are based on the decision-theoretic or syntactic approaches to pattern recognition. In this chapter we shall consider much simpler techniques, which are usually implemented by relatively inexpensive, dedicated hardware. The algorithms are primarily heuristic and are usually specific to a given wavelet (say the QRS complex in the ECG).

A more general approach is the one where the approximate knowledge on the wavelet is used to generate a template, which is some average wavelet determined by the a priori knowledge on the signal. The signal is then continuously matched with the template by means of correlation, matched filtering, or other pattern recognition techniques. The method is general in the sense that one can use the same algorithm for the detection of any wavelet by just replacing the template. The choice bet...