Simplified Wavelet Filter Implementation for Real-Time Signal Recognition

This article describes a simple implementation of wavelet filters for recognizing the shape of pulsed signals in real time. The proposed approach allows implementing a multi-level wavelet transform both in a discrete form based on microcontrollers and in a discrete-analog form using simple operations of addition and multiplication by a constant. The article describes a simple algorithm for calculating wavelet coefficients, estimates its speed and accuracy for performing calculations with a fixed point. The paper presents the calculation of the required number of memory cells for the implementation of the modified algorithm and shows that with a small filter length, some memory savings are achieved compared to the classical implementation. The article provides a simple implementation of an analog adder with accumulation of the result and proposes a functional diagram of a wavelet filter based on unified modules. The paper considers a practical example of the recognition of pulse signals based on the working alphabet of two classification criteria, which are approximating and detailing wavelet coefficients.