Beyond Oversmoothing: Evaluating DDPM and MSE for Scalable Speech Synthesis in ASR

• URL: http://arxiv.org/abs/2410.12279v1
• Date: Wed, 16 Oct 2024 06:35:56 GMT
• Title: Beyond Oversmoothing: Evaluating DDPM and MSE for Scalable Speech Synthesis in ASR
• Authors: Christoph Minixhofer, Ondrej Klejch, Peter Bell,
• Abstract summary: We compare Denoising Diffusion Probabilistic Models (DDPM) to Mean Squared Error (MSE) based models for TTS, when used for ASR model training. We find that for a given model size, DDPM can make better use of more data, and a more diverse set of speakers, than MSE models. We achieve the best reported ratio between real and synthetic speech WER to date (1.46), but also find that a large gap remains.
• Score: 13.307889110301502
• License:
• Abstract: Synthetically generated speech has rapidly approached human levels of naturalness. However, the paradox remains that ASR systems, when trained on TTS output that is judged as natural by humans, continue to perform badly on real speech. In this work, we explore whether this phenomenon is due to the oversmoothing behaviour of models commonly used in TTS, with a particular focus on the behaviour of TTS-for-ASR as the amount of TTS training data is scaled up. We systematically compare Denoising Diffusion Probabilistic Models (DDPM) to Mean Squared Error (MSE) based models for TTS, when used for ASR model training. We test the scalability of the two approaches, varying both the number hours, and the number of different speakers. We find that for a given model size, DDPM can make better use of more data, and a more diverse set of speakers, than MSE models. We achieve the best reported ratio between real and synthetic speech WER to date (1.46), but also find that a large gap remains.

Related papers

• SimpleSpeech 2: Towards Simple and Efficient Text-to-Speech with Flow-based Scalar Latent Transformer Diffusion Models [64.40250409933752]
  We build upon our previous publication by implementing a simple and efficient non-autoregressive (NAR) TTS framework, termed SimpleSpeech 2. SimpleSpeech 2 effectively combines the strengths of both autoregressive (AR) and non-autoregressive (NAR) methods. We show a significant improvement in generation performance and generation speed compared to our previous work and other state-of-the-art (SOTA) large-scale TTS models.
  arXiv  Detail & Related papers  (2024-08-25T17:07:39Z)
• Can We Achieve High-quality Direct Speech-to-Speech Translation without Parallel Speech Data? [49.42189569058647]
  Two-pass direct speech-to-speech translation (S2ST) models decompose the task into speech-to-text translation (S2TT) and text-to-speech (TTS) In this paper, we introduce a composite S2ST model named ComSpeech, which can seamlessly integrate any pretrained S2TT and TTS models into a direct S2ST model. We also propose a novel training method ComSpeech-ZS that solely utilizes S2TT and TTS data.
  arXiv  Detail & Related papers  (2024-06-11T14:17:12Z)
• EM-TTS: Efficiently Trained Low-Resource Mongolian Lightweight Text-to-Speech [4.91849983180793]
  We propose a lightweight Text-to-Speech (TTS) system based on deep convolutional neural networks. Our model consists of two stages: Text2Spectrum and SSRN. Experiments show that our model can reduce the training time and parameters while ensuring the quality and naturalness of the synthesized speech.
  arXiv  Detail & Related papers  (2024-03-13T01:27:57Z)
• Mega-TTS: Zero-Shot Text-to-Speech at Scale with Intrinsic Inductive Bias [71.94109664001952]
  Mega-TTS is a novel zero-shot TTS system that is trained with large-scale wild data. We show that Mega-TTS surpasses state-of-the-art TTS systems on zero-shot TTS speech editing, and cross-lingual TTS tasks.
  arXiv  Detail & Related papers  (2023-06-06T08:54:49Z)
• Unsupervised Data Selection for TTS: Using Arabic Broadcast News as a Case Study [44.07589545984369]
  We propose a fully unsupervised method for building TTS, including automatic data selection and pre-training/fine-tuning strategies. We show how careful selection of data, yet smaller amounts, can improve the efficiency of TTS system. Our objective evaluation shows 3.9% character error rate (CER), while the groundtruth has 1.3% CER.
  arXiv  Detail & Related papers  (2023-01-22T10:41:58Z)
• EPIC TTS Models: Empirical Pruning Investigations Characterizing Text-To-Speech Models [26.462819114575172]
  This work compares sparsity paradigms in text-to-speech synthesis. It is the first work that compares sparsity paradigms in text-to-speech synthesis.
  arXiv  Detail & Related papers  (2022-09-22T09:47:25Z)
• DDKtor: Automatic Diadochokinetic Speech Analysis [13.68342426889044]
  This paper presents two deep neural network models that automatically segment consonants and vowels from unannotated, untranscribed speech. Results on a young healthy individuals dataset show that our LSTM model outperforms the current state-of-the-art systems. The LSTM model also presents comparable results to trained human annotators when evaluated on unseen older individuals with Parkinson's Disease dataset.
  arXiv  Detail & Related papers  (2022-06-29T13:34:03Z)
• Sequence-level self-learning with multiple hypotheses [53.04725240411895]
  We develop new self-learning techniques with an attention-based sequence-to-sequence (seq2seq) model for automatic speech recognition (ASR) In contrast to conventional unsupervised learning approaches, we adopt the emphmulti-task learning (MTL) framework. Our experiment results show that our method can reduce the WER on the British speech data from 14.55% to 10.36% compared to the baseline model trained with the US English data only.
  arXiv  Detail & Related papers  (2021-12-10T20:47:58Z)
• On the Interplay Between Sparsity, Naturalness, Intelligibility, and Prosody in Speech Synthesis [102.80458458550999]
  We investigate the tradeoffs between sparstiy and its subsequent effects on synthetic speech. Our findings suggest that not only are end-to-end TTS models highly prunable, but also, perhaps surprisingly, pruned TTS models can produce synthetic speech with equal or higher naturalness and intelligibility.
  arXiv  Detail & Related papers  (2021-10-04T02:03:28Z)
• Learning to Maximize Speech Quality Directly Using MOS Prediction for Neural Text-to-Speech [15.796199345773873]
  We propose a novel method to improve speech quality by training a TTS model under the supervision of perceptual loss. We first pre-train a mean opinion score (MOS) prediction model and then train a TTS model to maximize the MOS of synthesized speech. The proposed method can be applied independently regardless of the TTS model architecture or the cause of speech quality degradation.
  arXiv  Detail & Related papers  (2020-11-02T18:13:48Z)
• Joint Contextual Modeling for ASR Correction and Language Understanding [60.230013453699975]
  We propose multi-task neural approaches to perform contextual language correction on ASR outputs jointly with language understanding (LU) We show that the error rates of off the shelf ASR and following LU systems can be reduced significantly by 14% relative with joint models trained using small amounts of in-domain data.
  arXiv  Detail & Related papers  (2020-01-28T22:09:25Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.