Influence of various asymmetrical contextual factors for TTS in a low resource language

Abstract

The generalized statistical framework of Hidden Markov Model (HMM) has been successfully applied from the field of speech recognition to speech synthesis. In this paper, we have applied HMM-based Speech Synthesis (HTS) method to Gujarati (one of the official languages of India). Adaption and evaluation of HTS for Gujarati language has been done here. In addition, to understand the influence of asymmetrical contextual factors on quality of synthesized speech, we have conducted series of experiments. Evaluation of different HTS built for Gujarati speech using various asymmetrical contextual factors is done in terms of naturalness and speech intelligibility. From the experimental results, it is evident that when more weightage is given to left phoneme in asymmetrical contextual factor, HTS performance improves compared to conventional symmetrical contextual factors for both triphone and pentaphone case. Furthermore, we achieved best performance for Gujarati HTS with left-left-left-centre-right (i.e., LLLCR) contextual factors.

Effectiveness of Generative Adversarial Network for Non-Audible Murmur-to-Whisper Speech Conversion

“…Since speech is a sequential data, extracting the contextual features from the speech, captures the local features (including coarticulation) and preserves the crucial harmonics [31,32]. In speech perception, it has been shown that the surrounding acoustic context, impacts the human perception [33][34][35].…”

“…Inspired by the study reported in [32], we also analyze the importance of training models by taking an asymmetric contextual frames as an input to the network (Panel II in Figure 2). No significant variations could be observed in terms of MCD scores for GAN-based systems (as shown in Figure 2 (c)), whereas significant improvement in the performance of the GAN-based system over DNN-based system, is observed in terms of PESQ score (as shown in Figure 2 (d)) (notably 4LC1R).…”

Effectiveness of Dynamic Features in INCA and Temporal Context-INCA

“…To overcome this issue, Temporal-Context (TC) INCA algorithm was proposed [22], which tries to incorporate the contextual information. Furthermore, since speech is a sequential data, extracting the contextual features from the speech, captures the local features (including coarticulation) and preserves the crucial harmonics [23,24]. It is well known in the speech literature that the surrounding acoustic context affects the human speech perception [25][26][27].…”