Aperiodicity extraction and control using mixed mode excitation and group delay manipulation for a high quality speech analysis, modification and synthesis system STRAIGHT

By Hideki Kawahara, J. Estill, O. Fujimura. 2001

322 citations recorded by the source corpus.

Subfields: Natural Language Processing, Speech & Audio

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

1. Junichi Yamagishi
2. Graham Neubig
3. Tie-Yan Liu
4. Tao Qin
5. Xu Tan
6. Yoshua Bengio
7. Kyunghyun Cho
8. Heiga Zen

In Speech & Audio

1. Junichi Yamagishi
2. Xu Tan
3. Tao Qin
4. Tomoki Toda
5. Frank K. Soong
6. Tie-Yan Liu
7. Heiga Zen
8. Sheng Zhao

Citations in the corpus

• WaveNet: A Generative Model for Raw Audio (2016, background): “It often includes cepstra (Imai & Furuichi, 1988) or line spectral pairs (Itakura, 1975), which represent vocal tract transfer function, and fundamental frequency ( F 0 ) and aperiodicity (Kawahara et al., 2001), which represent characteristics of vocal source excitation signals.”
• A Survey on Neural Speech Synthesis (2021, methodology): “In vocoder analysis, it analyzes the speech and gets acoustic features such as mel-cepstral coefficients [79], band aperiodicity [153, 154] and F0.”
• Postfilters to Modify the Modulation Spectrum for Statistical Parametric Speech Synthesis (2016, methodology): “The 0th-through-24th mel-cepstral coef-ficients were extracted as spectral parameters and log-scaled F 0 and 5 band-aperiodicity [62], [63] were extracted as excitation parameters.”
• A Survey on Audio Diffusion Models: Text To Speech Synthesis and Enhancement in Generative AI (2023, unknown): “Common acoustic features include mel-cepstral coefficients [18], Mel-generalized coefficients [114], F0 [34] and band aperiodicity [33].”
• Generative Adversarial Network-Based Glottal Waveform Model for Statistical Parametric Speech Synthesis (2017, methodology): “Vocoders used in SPSS can be grouped into three main categories: mixed/impulse excited vocoders (e.g. STRAIGHT [8, 9]), glottal vocoders (e.g. Glot-tHMM [10] and GlottDNN [11]), and sinusoidal vocoders (e.g. quasi harmonic model [12, 13]).”
• Directly modeling speech waveforms by neural networks for statistical parametric speech synthesis (2015, background): “cepstra [2], line spectrum pairs [3], fundamental frequency, and aperiodicity [4].”
• Current trends in multilingual speech processing (2011, methodology): “It is also evident that the dominant source-filter model employed in HMM-based TTS (Koishida et al 1994; Kawahara et al 2001) may not be 10See respectively: http://www.darpa.mil/ipto/programs/transtac/transtac.asp; http://www.tc-star.org; http://www.darpa. mil/ipto/programs/gale/gale.asp ideal for…”
• Excitation source analysis for high-quality speech manipulation systems based on an interference-free representation of group delay with minimum phase response compensation (2014, methodology): “The most crucial issue is excitation source representations, especially non-periodic components [9, 10, 11].”
• Speech Enhancement of Noisy and Reverberant Speech for Text-to-Speech (2018, methodology): “The shape of these two windows are designed in order to smooth peaks and valleys respectively [38].”
• Towards Achieving Robust Universal Neural Vocoding (2018, methodology): “In the case of traditional vocoding [7, 8, 9, 10], approaches commonly relied on simplified models (e.”
• Statistical Parametric Speech Synthesis Incorporating Generative Adversarial Networks (2017, methodology): “The 0th-through-24th mel-cepstral coefficients were used as spectral parameters and F 0 and 5 band-aperiodicity [47], [48] were used as excitation parameters.”
• GlotNet—A Raw Waveform Model for the Glottal Excitation in Statistical Parametric Speech Synthesis (2019, methodology): “STRAIGHT [35], [44] is a classical source-filter vocoder”
• Robust Speaker-Adaptive HMM-Based Text-to-Speech Synthesis (2009, unknown): “Source Pulse/noise Mixed [16] Mixed Pulse/noise Mixed”
• Directly modeling voiced and unvoiced components in speech waveforms by neural networks (2016, background): “Index Terms— Statistical parametric speech synthesis; neural network; wavefom”
• Robust universal neural vocoding (2018, methodology): “In the case of traditional vocoding [6, 7, 8, 9], approaches commonly relied on simplified models (e.”
• PYIN: A fundamental frequency estimator using probabilistic threshold distributions (2014, background): “Several solutions to the problem of F0 estimation have been proposed in the area of speech processing [4, 5, 6, 7].”