We report a new method using a time delay neural network to transform acoustic emission (AE) waveforms into a time series of instantaneous frequency content and permutation entropy. This permits periods of noise to be distinguished from signals. The model is trained in sequential batches, using an automated process that steadily improves signal recognition as new data are added. The model was validated using AE data from rock deformation experiments, using Darley Dale sandstone in fully drained conditions at a confining pressure of 20 MPa (approximately 800 m simulated depth). The model is initially trained by manual picking of five high‐amplitude waveforms randomly selected from the dataset (experiment). This is followed by semisupervised training on a subset of 300 waveforms.