Jorge Wuth

ASR based pronunciation evaluation with automatically generated competing vocabulary and classifier fusion

In this paper the application of automatic speech recognition (ASR) technology in CAPT (Computer Aided Pronunciation Training) is addressed. A method to automatically generate the competitive lexicon, required by an ASR engine to compare the pronunciation of a target word with its correct and wrong phonetic realization, is presented. In order to enable the efficient deployment of CAPT applications, the generation of this competitive lexicon does not require any human assistance or a priori information of mother language dependent errors. The method presented here leads to averaged subjective-objective score correlation equal to 0.82 and 0.75 depending on the task. Index Terms: second language learning, computer aided pronunciation training, speech recognition and competing vocabulary

Improved signal detection algorithms for unevenly sampled data. Six signals in the radial velocity data for GJ876

The hunt for Earth analogue planets orbiting Sun-like stars has forced the introduction of novel methods to detect signals at, or below, the level of the intrinsic noise of the observations. We present a new global periodogram method that returns more information than the classic Lomb-Scargle periodogram method for radial velocity signal detection. Our method uses the Minimum Mean Squared Error as a framework to determine the optimal number of genuine signals present in a radial velocity timeseries using a global search algorithm, meaning we can discard noise spikes from the data before follow-up analysis. This method also allows us to determine the phase and amplitude of the signals we detect, meaning we can track these quantities as a function of time to test if the signals are stationary or non-stationary. We apply our method to the radial velocity data for GJ876 as a test system to highlight how the phase information can be used to select against non-stationary sources of detected signals in radial velocity data, such as rotational modulation of star spots. Analysis of this system yields two new statistically significant signals in the combined Keck and HARPS velocities with periods of 10 and 15 days. Although a planet with a period of 15 days would relate to a Laplace resonant chain configuration with three of the other planets (8:4:2:1), we stress that follow-up dynamical analyses are needed to test the reliability of such a six planet system.

Maximum Entropy-Based Reinforcement Learning Using a Confidence Measure in Speech Recognition for Telephone Speech

In this paper, a novel confidence-based reinforcement learning (RL) scheme to correct observation log-likelihoods and to address the problem of unsupervised compensation with limited estimation data is proposed. A two-step Viterbi decoding is presented which estimates a correction factor for the observation log-likelihoods that makes the recognized and neighboring HMMs more or less likely by using a confidence score. If regions in the output delivered by the recognizer exhibit low confidence scores, the second Viterbi decoding will tend to focus the search on neighboring models. In contrast, if recognized regions exhibit high confidence scores, the second Viterbi decoding will tend to retain the recognition output obtained at the first step. The proposed RL mechanism is modeled as the linear combination of two metrics or information sources: the acoustic model log-likelihood and the logarithm of a confidence metric. A criterion based on incremental conditional entropy maximization to optimize a linear combination of metrics or information sources online is also presented. The method requires only one utterance, as short as 0.7 s, and can lead to significant reductions in word error rate (WER) between 3% and 18%, depending on the task, training-testing conditions, and method used to optimize the proposed RL scheme. In contrast to ordinary feature compensation and model parameter adaptation methods, the confidence-based RL method takes place in the frame log-likelihood domain. Consequently, as shown in the results presented here, it is complementary to feature compensation and to model adaptation techniques.

DNN-HMM based Automatic Speech Recognition for HRI Scenarios

In this paper, we propose to replace the classical black box integration of automatic speech recognition technology in HRI applications with the incorporation of the HRI environment representation and modeling, and the robot and user states and contexts. Accordingly, this paper focuses on the environment representation and modeling by training a deep neural network-hidden Markov model based automatic speech recognition engine combining clean utterances with the acoustic-channel responses and noise that were obtained from an HRI testbed built with a PR2 mobile manipulation robot. This method avoids recording a training database in all the possible acoustic environments given an HRI scenario. Moreover, different speech recognition testing conditions were produced by recording two types of acoustics sources, i.e. a loudspeaker and human speakers, using a Microsoft Kinect mounted on top of the PR2 robot, while performing head rotations and movements towards and away from the fixed sources. In this generic HRI scenario, the resulting automatic speech recognition engine provided a word error rate that is at least 26% and 38% lower than publicly available speech recognition APIs with the playback (i.e. loudspeaker) and human testing databases, respectively, with a limited amount of training data.