Flexible decision support system for ultrasound evaluation of fiber–metal laminates implemented in a DSP

• A comparative study on different feature extraction techniques for classification of ultrasound evaluation signals was presented.
• Principal component analysis is used for feature selection providing a compact set of decorrelated and discriminative features.
• Different classifier designs were evaluated.
• A flexible embedded implementation of the proposed system in a digital signal processor was presented.
• The digital system was evaluated considering the computational accuracy and the execution time.

Ultrasound testing has been widely applied for material characterization. The method accuracy usually relies on operator experience, considering this, an automatic decision support system may contribute to increase the evaluation efficiency. This paper presented an embedded electronic system for decision support in ultrasound evaluation of fiber–metal laminate composites. The proposed system comprised analog to digital conversion and digital signal processing algorithms. Discrete Fourier, wavelet and cosine transforms were used for feature extraction and principal component analysis was applied for efficient feature selection. The automatic classification was performed using an artificial neural network. The results demonstrated that it was possible to produce, in a short time latency, high-quality decision support information for two different types of test objects.