Search

Fountain code can significantly increase eavesdroppers’ untranslated efficiency in the wireless communication eavesdropping channel. The secondary LT coding anti-eavesdropping scheme with fountain code degree-1 is the subject of a theoretical investigation in this paper. The fact that its channel security capacity is greater than that of traditional LT code is first deduced from an information-theoretic standpoint, and the impact of source symbol length on decoding complexity and decoding overhead is then examined. The experimental results show that, compared with the traditional anti-eavesdropping twice fountain code, selecting long source symbols for double LT coding, when the main channel is better than the eavesdropping channel, can ensure that the eavesdropper has a higher untranslated efficiency, and can effectively reduce the fountain code decoding complexity and the number of encoded symbols sent by the source to improve the efficiency of information transmission.

The prediction of pause fillers plays a crucial role in enhancing the naturalness of synthesized speech. In recent years, neural networks, including LSTM, BERT, and XLNet, have been employed for pause fillers prediction modules. However, these methods have exhibited relatively lower accuracy in predicting pause fillers. This paper introduces the utilization of the RoBERTa model for predicting Chinese pause fillers and presents a novel approach to training the RoBERTa model, effectively enhancing the accuracy of Chinese pause fillers prediction. Our proposed approach involves categorizing text from different speakers into four distinct style groups based on the frequency and position of Chinese pause fillers. The RoBERTa model is trained on these four groups of data, which incorporate different styles of fillers, thereby ensuring a more natural synthesis of speech. The Chinese pause fillers prediction module is evaluated on systems such as Parallel Tacotron2, FastPitch, and Deep Voice3, achieving a notable 26.7% improvement in word-level prediction accuracy compared to the BERT model, along with a 14% enhancement in position-level prediction accuracy. This substantial improvement results in a significant enhancement of the naturalness of the generated speech.