Micro-emboli detection for patients with high risk of strokes has been performed with transcranial Doppler (TCD) systems since 1969. As a consequence, instrumentation of TCD systems progressed with the introduction of multigate systems, power mode systems and robotized probes, to name but a few. These new types of TCD have increase the chance of robust detections of quite big micro-emboli and at the same time increased the efficiency of artefact rejection. For a couple of years now, it is now possible to prevent cerebrovascular accidents (CVA) by detecting very small micro-emboli, the latter being precursor signs of strokes. For this sake, a new generation of transcranial Doppler (TCD) systems (holter) is used to record examinations of long duration. In an attempt to detect the smallest possible micro-emboli, offline softwares based on recent signal processing techniques complete advantageously these holter systems. In this communication, an overview of fifty years of research developments in embolic signal processing is proposed. What is interesting during this adventure of a half century is that detection methods were inspired as signal processing discoveries coming from speech processing to econometric. With the advent of the artificial intelligence, new challenges are being drawn up.