Evaluating ASR confidence score performance in different noisy conditions
نویسندگان
چکیده
This paper deals with the impact of a signal enhancement method on speech recognition confidence scores, in the presence of noise. The speech recognition confidence score reflects the reliability of correctness of the recogniser’s output. This is important since errors in the presence of noise are more frequent and tend to make applications, such as spoken dialogue systems, too cumbersome to use. By recognising enhanced signals not only the word error rate decreases but also the word-based confidence scores decline. Therefore, the role of speech recognition in spoken dialogue systems weakens. The experimentation involves input signal corrupted by coloured noise and cocktail party noise with varying Signal-to-Noise Ratio. A non-linear spectral subtraction method (NSS) will be used in conjunction with the Continuous Speech Recognition system developed for the ERMIS project (IST-2000-29319) to quantify the impact of speech enhancement on confidence score accuracy.
منابع مشابه
Reconnaissance d'ordres domotiques en conditions bruitées pour l'assistance à domicile (Recognition of Voice Commands by Multisource ASR and Noise Cancellation in a Smart Home Environment) [in French]
In this paper, we present a multisource ASR system to detect home automation orders in various everyday listening conditions in a realistic home. The system is based on a state of the art noise cancellation stage that feeds recently introduced ASR techniques. The evaluation was conducted on a realistic noisy dataset acquired in a smart home where a microphone was placed near the noise source an...
متن کاملDeep neural network based spectral feature mapping for robust speech recognition
Automatic speech recognition (ASR) systems suffer from performance degradation under noisy and reverberant conditions. In this work, we explore a deep neural network (DNN) based approach for spectral feature mapping from corrupted speech to clean speech. The DNN based mapping substantially reduces interference and produces estimated clean spectral features for ASR training and decoding. We expe...
متن کاملParameters Optimization for Improving ASR Performance in Adverse Real World Noisy Environmental Conditions
From the existing research it has been observed that many techniques and methodologies are available for performing every step of Automatic Speech Recognition (ASR) system, but the performance (Minimization of Word Error Recognition-WER and Maximization of Word Accuracy RateWAR) of the methodology is not dependent on the only technique applied in that method. The research work indicates that, p...
متن کاملNoise adaptation for robust AURORA 2 noisy digit recognition using statistical data mapping
The mismatch between system training and operating conditions often has negative influences on automatic speech recognition (ASR) systems. Noise in the operating environments is commonly encountered. ASR model adaptation is an important way to enhance the system performance in noisy environments. This paper proposes a feature-based statistical data mapping (SDM) approach for robust noisy digit ...
متن کاملMultiple Timescale Feature Combination Towards Robust Speech Recognition
While a lot of progress has been made during the last years in the field of Automatic Speech recognition (ASR), one of the main remaining problems is that of robustness. Typically, state-of-the-art ASR systems work very efficiently in well-defined environments, e.g. for clean speech or known noise conditions. However, their performance degrades drastically under different conditions. Many appro...
متن کامل