Black-box adversarial attacks through speech distortion for speech emotion recognition
Black-box attack on three speech emotion recognition models First, the data set is transformed into the spectral envelope in the speech distortion, and then the transformed speech data set is identified in three models (LSTM/GRAPH/MAA) to check the attack effect