Shijun Wang

Thumbnail Image
Last Name
Wang
First name
Shijun
Email
shijun.wang@unisg.ch
Phone
+41 71 224 73 70

Settings

Now showing 1 - 2 of 2
  • Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    FINE-GRAINED EMOTIONAL CONTROL OF TEXT-TO-SPEECH: LEARNING TO RANK INTER-AND INTRA-CLASS EMOTION INTENSITIES
    (2023-03-11)
    Wang, Shijun  
    ;
    Guðnason, Jón
    ;
    Borth, Damian  
    State-of-the-art Text-To-Speech (TTS) models are capable of producing high-quality speech. The generated speech, however, is usually neutral in emotional expression, whereas very often one would want fine-grained emotional control of words or phonemes. Although still challenging, the first TTS models have been recently proposed that are able to control voice by manually assigning emotion intensity. Unfortunately, due to the neglect of intra-class distance, the intensity differences are often unrecognizable. In this paper, we propose a fine-grained controllable emotional TTS, that considers both inter-and intra-class distances and be able to synthesize speech with recognizable intensity difference. Our subjective and objective experiments demonstrate that our model exceeds two state-of-the-art controllable TTS models for controllability, emotion expressiveness and naturalness.
    Type:conference paper
    URL:https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10097118
    DOI:10.1109/ICASSP43922.2022
    URI:https://www.alexandria.unisg.ch/handle/20.500.14171/118714
  • Thumbnail Image
    Some of the metrics are blocked by your 
    consent settings
    Item type:Publication,
    Learning Emotional Representations from Imbalanced Speech Data for Speech Emotion Recognition and Emotional Text-to-Speech
    (2023-06-21)
    Wang, Shijun  
    ;
    Guðnason, Jón
    ;
    Borth, Damian  
    Effective speech emotional representations play a key role in Speech Emotion Recognition (SER) and Emotional Text-To-Speech (TTS) tasks. However, emotional speech samples are more difficult and expensive to acquire compared with Neutral style speech, which causes one issue that most related works unfortunately neglect: imbalanced datasets. Models might overfit to the majority Neutral class and fail to produce robust and effective emotional representations. In this paper, we propose an Emotion Extractor to address this issue. We use augmentation approaches to train the model and enable it to extract effective and generalizable emotional representations from imbalanced datasets. Our empirical results show that (1) for the SER task, the proposed Emotion Extractor surpasses the state-of-the-art baseline on three imbalanced datasets; (2) the produced representations from our Emotion Extractor benefit the TTS model, and enable it to synthesize more expressive speech.
    Type:conference paper
    DOI:10.21437/Interspeech.2023
    URI:https://www.alexandria.unisg.ch/handle/20.500.14171/118715