Application of Artificial Intelligence TTS Technology in Simulating the Sound Characteristics of Professional Broadcasters

To enhance the simulation accuracy of professional announcers’ vocal traits and produce more natural-sounding synthetic speech, researchers conducted applied studies on artificial intelligence Text-to-Speech (TTS) technology. These studies focused on replicating the distinctive voice characteristics of professional broadcasters through advanced AI modeling techniques. The research aimed to achieve more human-like speech synthesis by refining acoustic feature extraction, prosody modeling, and emotional expression parameters in TTS systems. This paper first regularizes the Chinese text data based on the Bi LSTM neural network and then introduces the Empirical Mode Decomposition theory to complete the professional announcer voiceprint feature extraction. Finally, taking the regularized Chinese data and the extracted professional announcer voiceprint feature as the input, the professional announcer voice simulation model is constructed based on the improved FastSpeech2. The experimental results show that the similarity between the speech voiceprint features obtained by the proposed method and those of professional broadcasters is higher than 87.6%, with an Receive Frequency Test (RFT) value of 0.155. The application effect is better than comparative methods, and it has greater application value.