An adaptive algorithm for mel-cepstral analysis of speech

By Toshiaki Fukada, Keiichi Tokuda, Takao Kobayashi, Satoshi Imai. 1992

How researchers see this paper

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

Citations in the corpus

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.”
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].”
The HMM-based speech synthesis system (HTS) version 2.0 (2007, background): “The main difference is that both spectrum (mel-cepstral coefficients [11], and their dynamic features) and excitation (logarithmic fundamental frequencies (logF0) and its dynamic features) parameters are extracted from a speech database and modeled by contextdependent HMMs (phonetic, linguistic, and”
Emotional voice conversion using deep neural networks with MCC and F0 features (2016, background): “Meanwhile, we respectively use the MCC features of the source and target voice as the input-layer data and output-layer data for DBNs. Fig.”
Integration of Spectral Feature Extraction and Modeling for HMM-Based Speech Synthesis (2014, methodology): “For feature extraction, a statistical parametric mel-cepstral analysis [11], [12] has been widely used.”
Statistical parametric speech synthesis using deep neural networks (2013, methodology): “The speech data was downsampled from 48 kHz to 16 kHz sampling, then 40 Mel-cepstral coefficients [34], logarithmic fundamental frequency (log F0) values, and 5-band aperiodicities (0–1, 1–2, 2–4, 4–6, 6–8 kHz) [33] were extracted every 5 ms.”
Unidirectional long short-term memory recurrent neural network with recurrent output layer for low-latency speech synthesis (2015, methodology): “The speech data was downsampled from 48 kHz to 16 kHz, then 40 mel-cepstral coefficients [30], logarithmic fundamental frequency (logF0) values, and 5-band aperiodicities (0–1, 1–2, 2–4, 4–6, 6–8 kHz) [29] were extracted every 5 ms.”
Simultaneous Modeling of Spectrum, Pitch and Duration in HMM-Based Speech Synthesis (1999, methodology): “Sequences of mel-cepstral coefﬁcient vector, which are obtained from speech database using a mel-cepstral analysis technique [5], are modeled by continuous density HMMs.”
Statistical Parametric Speech Synthesis (2007, background): “This process is very similar to that for speech recognition, the main difference being that both spectrum (e.g., mel-cepstral co-ef ﬁ cients (Fukada et al., 1992) and their dynamic features) and excitation (e.g., and its dynamic features) parameters are extracted from a database of natural speech…”
Speech Synthesis Based on Hidden Markov Models (2013, methodology): “Finally, a speech waveform is resynthesized directly from the generated spectral and excitation parameters by using a speech synthesis filter, such as the mel-log spectral approximation filter [42] for melcepstral coefficients and all-pole filter for linear-predictionbased spectral parameter coeffic”
Speech parameter generation algorithms for HMM-based speech synthesis (2000, methodology): “6-ms Blackman window with a 5ms shift, and then mel-cepstral coefficients were obtained by a mel-cepstral analysis technique [18].”
An HMM-based speech synthesis system applied to English (2002, methodology): “Thus we use a mel-cepstral analysis technique [7] which enables speech to be re-synthesized directly from the mel-cepstral coefficients using the MLSA (Mel Log Spectrum Approximation) filter [7](1).”
A Deep Generative Architecture for Postfiltering in Statistical Parametric Speech Synthesis (2015, methodology): “The spectrum of a speech frame is parametrized by the mel-cepstrum [22], resulting in a vector of length , which is the order of the cepstral analysis.”
DNN-based stochastic postfilter for HMM-based speech synthesis (2014, methodology): “In this work, the spectrum of a speech frame is parametrised by the mel-cepstrum [14], resulting in a vector c = [c1, c2, · · · , cM ] of length M , which is the order of the cepstral analysis.”
A Speech Parameter Generation Algorithm Considering Global Variance for HMM-Based Speech Synthesis (2007, methodology): “Then, a speech waveform was synthesized with the MLSA ﬁlter [13] based on the generated mel-cepstra.”
Continuous Stochastic Feature Mapping Based on Trajectory HMMs (2011, methodology): “Speech signals were sampled at a rate of 16 kHz and windowed by a 25-ms Blackman window with 5 ms of shift, and then 13th-order mel-cepstral coefficients were obtained by using a mel-cepstral analysis technique [8].”
ASVspoof 2015: the first automatic speaker verification spoofing and countermeasures challenge (2015, methodology): “As illustrated in Table 2, S1, S2, S3, and S4, all use the same STRAIGHT vocoder [25] for synthesis, whereas S5 uses an MLSA vocoder [23].”