Odile Mella

A wavelet-based parameterization for speech/music discrimination

This paper addresses the problem of parameterization for speech/music discrimination. The current successful parameterization based on cepstral coefficients uses the Fourier transformation (FT), which is well adapted for stationary signals. In order to take into account the non-stationarity of music/speech signals, this work proposes to study wavelet-based signal decomposition instead of FT. Three wavelet families and several numbers of vanishing moments have been evaluated. Different types of energy, calculated for each frequency band obtained from wavelet decomposition, are studied. Static, dynamic and long-term parameters were evaluated. The proposed parameterization are integrated into two class/non-class classifiers: one for speech/non-speech, one for music/non-music. Different experiments on realistic corpora, including different styles of speech and music (Broadcast News, Entertainment, Scheirer), illustrate the performance of the proposed parameterization, especially for music/non-music discrimination. Our parameterization yielded a significant reduction of the error rate. More than 30% relative improvement was obtained for the envisaged tasks compared to MFCC parameterization.

Optimizing the coverage of a speech database through a selection of representative speaker recordings

In the context of the Neologos French speech database creation project, we have defined a general methodology for the selection of representative speaker recordings. The selection aims at insuring a good coverage in terms of speaker variability while limiting the number of recorded speakers. This makes the resulting database both more adapted to the development of recently proposed multi-model methods and cheaper to collect. The presented methodology proposes to operate a selection by optimizing a quality criterion defined in a variety of speaker similarity modeling frameworks. The selection can be operated and validated with respect to a unique similarity criterion, using classical clustering methods such as Hierarchical or K-Medians clustering, or it can be operated and validated across several speaker similarity criteria, thanks to a newly developed clustering method called Focal Speakers Selection. In this framework, four different speaker similarity criteria are tested, and three different speaker clustering algorithms are compared. Results pertaining to the collection of the Neologos database are also discussed.

FRAME-SYNCHRONOUS AND LOCAL CONFIDENCE MEASURES FOR AUTOMATIC SPEECH RECOGNITION

In this paper, we introduce two new confidence measures for large vocabulary speech recognition systems. The major feature of these measures is that they can be computed without waiting for the end of the audio stream. We proposed two kinds of confidence measures: frame-synchronous and local. The frame-synchronous ones can be computed as soon as a frame is processed by the recognition engine and are based on a likelihood ratio. The local measures estimate a local posterior probability in the vicinity of the word to analyze. We evaluated our confidence measures within the framework of the automatic transcription of French broadcast news with the EER criterion. Our local measures achieved results very close to the best state-of-the-art measure (EER of 23% compared to 22.0%). We then conducted a preliminary experiment to assess the contribution of our confidence measure in improving the comprehension of an automatic transcription for the hearing impaired. We introduced several modalities to highlight words of low confidence in this transcription. We showed that these modalities used with our local confidence measure improved the comprehension of automatic transcription.

Qualitative investigation of the display of speech recognition results for communication with deaf people

Speech technologies provide ways of helping people with hearing loss by improving their autonomy. This study focuses on an application in French language which is developed in the collaborative project RAPSODIE in order to improve communication between a hearing person and a deaf or hard-of-hearing person. Our goal is to investigate different ways of displaying the speech recognition results which takes also into account the reliability of the recognized items. In this qualitative study, 10 persons have been interviewed to find the best way of displaying the speech transcription results. All the participants are deaf with different levels of hearing loss and various modes of communication.

Development of the Arabic Loria Automatic Speech Recognition system (ALASR) and its evaluation for Algerian dialect

Abstract This paper addresses the development of an Automatic Speech Recognition system for Modern Standard Arabic (MSA) and its extension to Algerian dialect. Algerian dialect is very different from Arabic dialects of the Middle-East, since it is highly influenced by the French language. In this article, we start by presenting the new automatic speech recognition named ALASR (Arabic Loria Automatic Speech Recognition) system. The acoustic model of ALASR is based on a DNN approach and the language model is a classical n-gram. Several options are investigated in this paper to find the best combination of models and parameters. ALASR achieves good results for MSA in terms of WER (14.02%), but it completely collapses on an Algerian dialect data set of 70 minutes (a WER of 89%). In order to take into account the impact of the French language, on the Algerian dialect, we combine in ALASR two acoustic models, the original one (MSA) and a French one trained on ESTER corpus. This solution has been adopted because no transcribed speech data for Algerian dialect are available. This combination leads to a substantial absolute reduction of the word error of 24%.

Frame-synchronous and local confidence measures for on-the-fly keyword spotting

This paper presents several new confidence measures for speech recognition applications. The major advantage of these measures is that they can be evaluated with only a part of the whole sentence. Two of these measures can be computed directly within the first step of the recognition process, synchronously with the decoding engine. Such measures are useful to drive the recognition process by modifying the likelihood score or to validate recognized words in on-the-fly applications as keyword spotting task and on-line automatic speech transcription for deaf people. Two kinds of results are given. Firstly, an EER evaluation on a French broadcast news corpus shows performance close to the batch version of these measures (23.9% against 23.8% of EER). Secondly, for the keyword spotting application, our best measure provides a decrease of the false-acceptation rate by 50% with only a decrease of the correct words by 5%.