## Spectral Envelopes and Inverse FFT Synthesis (1992)

Venue: | Proceedings of the 93rd Audio Engineering Society Convention |

Citations: | 29 - 2 self |

### BibTeX

@INPROCEEDINGS{Rodet92spectralenvelopes,

author = {X. Rodet and P. Depalle},

title = {Spectral Envelopes and Inverse FFT Synthesis},

booktitle = {Proceedings of the 93rd Audio Engineering Society Convention},

year = {1992}

}

### Years of Citing Articles

### OpenURL

### Abstract

We present a new additive synthesis method based on spectral envelopes and inverse Fast Fourier Transform (FFT -1 ). User control is facilitated by the use of spectral envelopes to describe the characteristics of the short term spectrum of the sound in terms of sinusoidal and noise components. Such characteristics can be given by users or obtained automatically from natural sounds. Use of the inverse FFT reduces the computation cost by a factor on the order of 15 compared to oscillators. We propose a low cost real-time synthesizer design allowing processing of recorded and live sounds, synthesis of instruments and synthesis of speech and the singing voice. Introduction Many musical sound signals may be described as a combination of a pseudoperiodic waveform and of colored noise [1]. The pseudo-periodic part of the signal can be viewed as a sum of sinusoidal components, named partials, with time-varying frequency and amplitude [2]. Such sinusoidal components are easily observed on ...

### Citations

102 | A system for sound analysis/transformation/synthesis based on a deterministic plus stochastic decomposition
- Serra
- 1989
(Show Context)
Citation Context ... 1992, San Francisco 5 noise signal. This is easy and inexpensive if the STFT has been computed and stored in a table before the beginning of the synthesis stage. There exist analysis methods [5],[8],=-=[9]-=- to separate the noise components from the sinusoidal ones, allowing the preparation of tables of noise component STS's. Implementation and performances The algorithm has been implemented first in UDI... |

23 |
Analysis of musical-instrument tones
- Risset, Mathews
- 1969
(Show Context)
Citation Context ...screte Fourier Transform. In consequence, some of the first attempts at sound synthesis were based on the method called additive synthesis, that is the summation of time-varying sinusoidal components =-=[3]-=-,[4]. This signal modeling approach inherits a rich history of signal processing techniques. As an example, we have developed methods to automatically analyze sounds in terms of partials and noise tha... |

17 |
A new analysis/synthesis system of musical signals using Prony’s method. Application to heavily damped percussive sounds
- Laroche
- 1989
(Show Context)
Citation Context ...f a pseudoperiodic waveform and of colored noise [1]. The pseudo-periodic part of the signal can be viewed as a sum of sinusoidal components, named partials, with time-varying frequency and amplitude =-=[2]-=-. Such sinusoidal components are easily observed on a spectral analysis display (Figure 1), as obtained, for instance, from a bank of filters or from a Discrete Fourier Transform. In consequence, some... |

8 |
A Modular System for Analysis, Processing and Synthesis of Sound Signals
- Depalle
- 1991
(Show Context)
Citation Context ...ng time-domain signal swl [n]. To better explain the method, let us consider first an analysis method familiar to many people, the analysis by Fast Fourier Transform such as used in the Phase Vocoder =-=[7]-=-. In this Spectral Envelopes and Inverse FFT Synthesis, X. Rodet & P. Depalle, AES 1992, San Francisco . Spectral Envelopes and Inverse FFT Synthesis, X. Rodet & P. Depalle, AES 1992, San Francisco 3 ... |

7 |
Tools for Rapid Prototyping of Music Sound Synthesis Algorithms and Control Strategies
- Freed
- 1992
(Show Context)
Citation Context ...reparation of tables of noise component STS's. Implementation and performances The algorithm has been implemented first in UDI [10] and in real time on the SUN-Mercury Workstation [11] and SGI Indigo =-=[12]-=-. In terms of cost, one of the critical part of the FFT -1 algorithm is the construction of a STS S. For each partial, the critical part consists of K iterations of the form: a j .e if j .W[f j-k]. Fo... |

7 |
Speech analysis and synthesis methods based on spectral envelopes and voiced/unvoiced functions
- Rodet, Depalle, et al.
- 1987
(Show Context)
Citation Context ...vidual breakpoint function becomes impossible in practice. Another argument against such breakpoint functions is as follows. In the case of the voice and of certain instruments, a source filter model =-=[13]-=- is an adequate representation of some of the behavior of the partials. Then the amplitude of a component is a function of its frequency, i.e. the transfer function of the filter [14]. That is, the am... |

3 |
Prototyping and interpolation of multiple musical timbres using principal components-based analysis
- Sandell, Martens
- 1992
(Show Context)
Citation Context ...te Fourier Transform. In consequence, some of the first attempts at sound synthesis were based on the method called additive synthesis, that is the summation of time-varying sinusoidal components [3],=-=[4]-=-. This signal modeling approach inherits a rich history of signal processing techniques. As an example, we have developed methods to automatically analyze sounds in terms of partials and noise that ca... |

3 |
Analyse des signaux sonores en termes de partiels et de bruit. Extraction automatique des trajets frquentiels par des Modles de Markov Cachs" Mmoire de DEA en Automatique et Traitement du Signal
- Garcia
- 1992
(Show Context)
Citation Context ...istory of signal processing techniques. As an example, we have developed methods to automatically analyze sounds in terms of partials and noise that can then be applied directly to additive synthesis =-=[5]-=-. In the sinusoidal model, harmonic or inharmonic partials are easy to characterize and easy to synthesize. Another interesting aspect of additive synthesis is the simplicity of the mapping of partial... |

2 |
Analysis and Synthesis Models for Musical Applications", IEEE Workshop on application of digital signal processing to audio and acoustics
- Rodet
- 1989
(Show Context)
Citation Context ...ds, synthesis of instruments and synthesis of speech and the singing voice. Introduction Many musical sound signals may be described as a combination of a pseudoperiodic waveform and of colored noise =-=[1]-=-. The pseudo-periodic part of the signal can be viewed as a sum of sinusoidal components, named partials, with time-varying frequency and amplitude [2]. Such sinusoidal components are easily observed ... |

2 |
Diphone Sound Synthesis
- Rodet, Depalle
(Show Context)
Citation Context ... why we describe amplitudes of partials by use of successive spectral envelopes defined at specific instants, for example the beginning and the end of the attack , sustain and decay of a note, etc....=-=[18]-=-. Then at any instant t, the spectral envelope to be used is obtained by interpolation between two successive envelopes. Let us also say that the specific instants can also be found automatically for ... |

1 |
Synthse additive par FTT inverse", Rapport Interne IRCAM
- Depalle
- 1990
(Show Context)
Citation Context ...r to produce large or narrow band limited random signals by combined amplitude and phase modulation, this has rarely be done. Inverse Fast Fourier Transform additive synthesis (FFT -1 ) In our method =-=[6]-=-, the computation of the partials is not done by a bank of oscillators but by an Inverse Fast Fourier Transform, that is a transformation of a short term spectrum (STS) Sl [k] into the corresponding t... |

1 |
Speech analysis/synthesis based on a sinusoidal representation
- Quartieri
- 1986
(Show Context)
Citation Context ... AES 1992, San Francisco 5 noise signal. This is easy and inexpensive if the STFT has been computed and stored in a table before the beginning of the synthesis stage. There exist analysis methods [5],=-=[8]-=-,[9] to separate the noise components from the sinusoidal ones, allowing the preparation of tables of noise component STS's. Implementation and performances The algorithm has been implemented first in... |

1 |
a Unified DSP Interface for Sound Signals Analysis and Synthesis
- Rodet, UDI
- 1992
(Show Context)
Citation Context ...to separate the noise components from the sinusoidal ones, allowing the preparation of tables of noise component STS's. Implementation and performances The algorithm has been implemented first in UDI =-=[10]-=- and in real time on the SUN-Mercury Workstation [11] and SGI Indigo [12]. In terms of cost, one of the critical part of the FFT -1 algorithm is the construction of a STS S. For each partial, the crit... |

1 |
A SUN-Mercury Workstation
- Eckel, Rodet, et al.
- 1987
(Show Context)
Citation Context ...ones, allowing the preparation of tables of noise component STS's. Implementation and performances The algorithm has been implemented first in UDI [10] and in real time on the SUN-Mercury Workstation =-=[11]-=- and SGI Indigo [12]. In terms of cost, one of the critical part of the FFT -1 algorithm is the construction of a STS S. For each partial, the critical part consists of K iterations of the form: a j .... |

1 | The CHANT Project
- Rodet, Potard, et al.
(Show Context)
Citation Context ...rce filter model [13] is an adequate representation of some of the behavior of the partials. Then the amplitude of a component is a function of its frequency, i.e. the transfer function of the filter =-=[14]-=-. That is, the amplitude a j can be obtained automatically by evaluating some spectral function (Figure 5), named spectral envelope e(f j ), at the frequency f j of the partial j. The amplitude variat... |

1 |
Use of LPC Spectral Estimation for Analysis, Processing and Synthesis
- Rodet, Depalle
- 1986
(Show Context)
Citation Context ...spectral tilt or spectral centroid changes known to be related to loudness and brilliance. Spectral envelopes can be obtained automatically by different methods among which Linear Prediction analysis =-=[15]-=-. If the amplitudes and frequencies of the partials are already known from sinusoidal analysis, the Generalized Discrete Cepstral analysis [16],[17], provides reliable envelopes, the smoothness of whi... |

1 |
Generalised Discrete Cepstral Estimation of Sound Signals
- Galas
- 1991
(Show Context)
Citation Context ...rent methods among which Linear Prediction analysis [15]. If the amplitudes and frequencies of the partials are already known from sinusoidal analysis, the Generalized Discrete Cepstral analysis [16],=-=[17]-=-, provides reliable envelopes, the smoothness of which can be adjusted according to the order as in classical cepstral analysis. From the previous arguments, it appears why we describe amplitudes of p... |

1 |
Dcodage Acoustico-Phontique: apport de la dcomposition Temporelle Gnralise et de transformations spectrales non-linaires", Thse de l'URA 820
- Montaci
- 1991
(Show Context)
Citation Context ...o successive envelopes. Let us also say that the specific instants can also be found automatically for the optimal reconstruction of the sound by linear interpolation between two successive envelopes =-=[19]-=-. No need to add that an amplitude spectral envelope is even better adapted for the description of noise components. Frequencies and phases also can be described by similar envelopes that we call gene... |

1 |
Speech Analysis ansd Synthesis Methods Based on Spectral Envelopes and Voiced/Unvoiced Functions
- Rodet, Depalle, et al.
(Show Context)
Citation Context ...vidual breakpoint function becomes impossible in practice. Another argument against such breakpoint functions is as follows. In the ease of the voice and of certain instruments, a source filter model =-=[13]-=- is an adequate representation of some of the behavior of the partials. Then the amplitude of a component is a function of its frequency, i.e. the transfer function of the filter [14]. That is, the am... |

1 |
D6codage Acoustico-Phon6tique: apport de la d6composition TemporelIe G6n6ralis6e et de transformations spectrales non-lin6aires", Th_se de rURA 820
- Montaci6
- 1991
(Show Context)
Citation Context ...o successive envelopes. Let us also say that the specific instants can also be found automatically for the optimal reconstruction of the sound by linear interpolation between two successive envelopes =-=[19]-=-. It should not be necessary to add that an amplitude spectral envelope is even better adapted for the description of noise components. Frequencies and phases also can be described by similar envelope... |