[2505.00409] Perceptual Implications of Automatic Anonymization in Pathological Speech

PDF view of the paper entitled Cognitive Effects to hide automatic identity in the pathological speech, written by Soroosh Tayebi Arasteh and 13 other books

PDF view

a summary:The techniques of not disclosing their automatic identity are necessary for the moral participation of pathological speech data, yet its cognitive consequences remain limited. We offer a comprehensive analysis that focuses on human being for unknown pathological speech, using an organized protocol that includes ten indigenous and indigenous German listeners with various linguistic, linguistic and artistic backgrounds. The listeners evaluated the unknown words of the identity of 180 speakers that extend to the lip, the roll, the speech imbalance, the transfer imbalance, the dysfunction, and health controls. Speech was hidden using modern automatic methods (equal error rates within 30-40 %). The listeners have completed the tasks of discrimination, such as Torring, quality classification tasks under zero conditions (individual exposure) and few conditions (repeated exposure). Discrimination accuracy was generally high (91 % zero shots; 93 % few shots), but varied due to disorder (repeated measures Anova: P = 0.007), ranging from 96 % (speech defect) to 86 % (dysphagia). His identity is constantly lowered from the perceived quality (from 83 % to 59 %, P <0.001), with patterns of deterioration special pathology (ANOVA in one direction: P = 0.005). The original listeners showed an unimportant trend towards the highest original speech classifications (Delta = 4 %, p = 0.199), but this difference was small after his identity was not disclosed (Delta = 1 %, p = 0.724). No great bias on the basis of sex. Cognitive results are not associated with automatic scales. The clarity has been linked to the quality of the original speech, but not after the identity is hidden. These results emphasize the need for strategies not to disclose its informed and listening identity, which maintains both the privacy and cognitive safety.