Comparative Evaluation of Listeners Perceptions on Synthetic Speech Quality

Quennie Joy B. Mesa, Kyung-Tae Kim and Jing-Jon Kim,

In this paper, we describe the perceptual and quality evaluation of synthetic speech generated by an HMM-based speech synthesizer for the Tagalog language using two groups of listeners: trained and untrained (naïve listeners). In addition, the paper analyzes and compares listeners’ quality perception of the synthetic speech. A total of 20 native speakers participated in the assessment process. Subjective listening tests were conducted to determine speech intelligibility and acceptability. Results on intelligibility tests show that untrained listeners find much difficulty in identifying words in semantically unstructured sentences as compared with trained listeners. However, no difference was recorded on intelligibility evaluation at phoneme level. Acceptability studies showed that both listeners appreciated and enjoyed using the synthetic speech which further proves that a neutral, stable, intelligible and acceptable Tagalog synthetic speech was developed.

Volume 11 | 11-Special Issue

Pages: 967-971

DOI: 10.5373/JARDCS/V11SP11/20193122