## A Model For Efficient Formant Estimation (1996)

Venue: | in ICASSP-96 |

Citations: | 12 - 1 self |

### BibTeX

@INPROCEEDINGS{Welling96amodel,

author = {L. Welling and H. Ney},

title = {A Model For Efficient Formant Estimation},

booktitle = {in ICASSP-96},

year = {1996},

pages = {797--800}

}

### OpenURL

### Abstract

This paper presents a new method for estimating formant frequencies. The formant model is based on a digital resonator. Each resonator represents a segment of the short--time power spectrum. The complete spectrum is modeled by a set of digital resonators connected in parallel. An algorithm based on dynamic programming produces both the model parameters and segment boundaries that optimally match the spectrum. The main results of this paper are: 1) Modeling formants by digital resonators allows a reliable estimation of formant frequencies. 2) Digital resonators can be used efficiently in connection with dynamic programming. 3) A recognition test with formant frequencies results in a string error rate of 4.8% on the adult corpus of the TI digit string database. 1. INTRODUCTION An efficient and compact representation of the time-- varying characteristics of speech offers potential benefits for speech recognition. Therefore a variety of approaches such as formant tracking [7, 4, 10], ar...

### Citations

178 |
Software for a Cascade/Parallel Formant Synthesizer
- Klatt
- 1980
(Show Context)
Citation Context ...ex--conjugate poles and not for real poles. Peak--picking techniques are vulnerable to merged formants and spurious peaks. The approach described in this paper avoids the above mentioned problems. In =-=[6]-=-, a set of digital formant resonators connected in parallel or in cascade has been proposed for speech synthesis. In this paper, we propose to use digital resonators for formant estimation. We model t... |

176 | Digital processing of speech signals - Rabiner, Schafer - 1978 |

21 |
Formant Tracking using Hidden Markov Models and Vector Quantization
- Kopec
- 1986
(Show Context)
Citation Context ...ON An efficient and compact representation of the time-- varying characteristics of speech offers potential benefits for speech recognition. Therefore a variety of approaches such as formant tracking =-=[7, 4, 10]-=-, articulatory models [9] and auditory models [5] have been explored. For formant tracking, methods based on linear predictor analysis (LPC) have received considerable attention. Root--finding algorit... |

18 |
Network-based connected digit recognition
- Bush, Kopec
- 1987
(Show Context)
Citation Context ...match the short--time power spectrum of a speech segment. We have performed recognition tests using formants on the TI digit string database. Formants have also been estimated on the same database in =-=[3]-=-. The paper is organized as follows. Section 2 defines the formant model. Section 3 describes the dynamic programming algorithm that produces the optimal set of segment boundaries. Section 4 contains ... |

17 |
Speaker dependent and independent speech recognition experiments with an auditory model
- Hunt, Lefebvre
- 1988
(Show Context)
Citation Context ... varying characteristics of speech offers potential benefits for speech recognition. Therefore a variety of approaches such as formant tracking [7, 4, 10], articulatory models [9] and auditory models =-=[5]-=- have been explored. For formant tracking, methods based on linear predictor analysis (LPC) have received considerable attention. Root--finding algorithms are employed to find the zeros of the LPC pol... |

17 |
Formant location from lpc analysis data
- Snell, Milinazzo
- 1993
(Show Context)
Citation Context ...ON An efficient and compact representation of the time-- varying characteristics of speech offers potential benefits for speech recognition. Therefore a variety of approaches such as formant tracking =-=[7, 4, 10]-=-, articulatory models [9] and auditory models [5] have been explored. For formant tracking, methods based on linear predictor analysis (LPC) have received considerable attention. Root--finding algorit... |

10 |
Globally optimising formant tracker using generalised centroids
- Crowe, Jack
- 1987
(Show Context)
Citation Context ...ON An efficient and compact representation of the time-- varying characteristics of speech offers potential benefits for speech recognition. Therefore a variety of approaches such as formant tracking =-=[7, 4, 10]-=-, articulatory models [9] and auditory models [5] have been explored. For formant tracking, methods based on linear predictor analysis (LPC) have received considerable attention. Root--finding algorit... |

10 | Deriving articulatory representations of speech
- Richards, Mason, et al.
- 1995
(Show Context)
Citation Context ...esentation of the time-- varying characteristics of speech offers potential benefits for speech recognition. Therefore a variety of approaches such as formant tracking [7, 4, 10], articulatory models =-=[9]-=- and auditory models [5] have been explored. For formant tracking, methods based on linear predictor analysis (LPC) have received considerable attention. Root--finding algorithms are employed to find ... |

10 | Connected Digit Recognition Using Statistical Template Matching
- Welling, Ney, et al.
- 1995
(Show Context)
Citation Context ...tic vectors for recognition experiments on the TI digit string database. The recognition system is based on hidden Markov models with continuous observation densities. Its characteristic features are =-=[11]-=-: 1) gender-- dependent word models for 11 English digits including 'oh' and gender--dependent silence models; 2) 357 states plus 1 state for silence per gender; 3) single Laplacian densities with sta... |

6 |
Sedgwick â€™A Method for Segmenting Acoustic Patterns, with Applications to Automatic Speech Recognitionâ€™, Transactions on Automatic Control
- Bridle, Nigel
(Show Context)
Citation Context .... DYNAMIC PROGRAMMING ALGORITHM The task is now to find the segment boundaries i 1 ; : : : ; i K \Gamma1 so that K X k=1 Emin (ik\Gamma1 + 1; i k ) is minimized (i0 = 0; i K = I). Dynamic programming =-=[1, 2]-=- provides an efficient solution. We introduce an auxiliary quantity F (k; i) which is defined as the error of the best segmentation of the frequency interval [1; i] into k segments. By decomposing the... |

2 |
An Introduction to the Theory of Dynamic
- Bellman
- 1953
(Show Context)
Citation Context .... DYNAMIC PROGRAMMING ALGORITHM The task is now to find the segment boundaries i 1 ; : : : ; i K \Gamma1 so that K X k=1 Emin (ik\Gamma1 + 1; i k ) is minimized (i0 = 0; i K = I). Dynamic programming =-=[1, 2]-=- provides an efficient solution. We introduce an auxiliary quantity F (k; i) which is defined as the error of the best segmentation of the frequency interval [1; i] into k segments. By decomposing the... |