Acoustic emission (EA) has proven to be very suitable for detecting and monitoring cracking of materials and structures. EA signals can be analyzedeither based on physical considerations (geophysics/seismology) or using their temporal and frequency characteristics. However, the multitude ofdefinitions related to the different parameters as well as the treatment methods make it necessary to develop a comparative analysis in the case of aheterogeneous material such as civil concrete. To this end, this contribution aims to study the microcracking of reinforced concrete T-beamssubjected to quasi-static mechanical tests. To do this, four-point bending tests, carried out at different travel speeds, were carried out in the presence ofa network of acoustic emission sensors. A comparison between the damage susceptibility of three definitions corresponding to the parameter b-valuewas carried out and supplemented by the evolution of the RA value and the mean frequency (AF) as a function of loading time. This work also showsthe use of the support-vector machine (SVM) method to define different areas of damage in the load-displacement curve. This work shows the limitations of this approach and proposes the use of an unsupervised learning approach to group EA data according to physical parameters as well astime/frequency parameters. Finally, this work discusses the advantages and limitations of the different methods and parameters used in relation to the micro/macro mechanisms at the origin of concrete cracking.