FIG. 13B is a block diagram illustrating the use of the neural network 1310 after it has been trained, according to an embodiment. An acoustic signal 1318 (not part of the training set used to train the neural network 1310) is received by the pre-processor 1304, which converts it into a spectrogram 1306. The trained neural network 1310 receives the spectrogram 1306, and outputs a set of haptic cues 1322 for each time slice of the spectrogram. These haptic cues 1322 indicate a pattern of haptic outputs for the cutaneous actuators of the haptic device 1324, but may not yet be in a format that can be used by the haptic device 1324 directly. For example, the haptic cues 1322 may be output as a set of percentage values, as described above with reference to FIG. 13A. Thus, the haptic cues 1322 may first be converted into actuator signals by a haptic signal generator 1323, before being transmitted to the cutaneous actuators 1324 to generate the corresponding haptic outputs.
