Search

The development of connected vehicle technology is beneficial to traffic operations at intersections. Automated vehicles are expected to emerge in the urban network gradually. The mixed traffic environment will exist for a long period. This paper presents a signal optimization model for an isolated intersection under the mixed traffic environment including connected and human-driven vehicles (CHVs) and connected and automated vehicles (CAVs). The problem is formulated as a mixed integer linear programming (MILP) model. Phase sequence and phase duration are optimized without the concept of cycle length for total vehicle delay minimization. Vehicle motion is analyzed and the passing states of CHVs and CAVs at stop bars are predicted to avoid the prior information of specific trajectory planning strategies of CAVs. A rolling horizon procedure is designed for the dynamic implementation of the proposed model with varying traffic conditions. The simulation results show that the proposed signal timing model can reduce average vehicle delay significantly compared to vehicle-actuated and fixed time control, especially with under-saturated traffic demand and more conflicting traffic streams. The sensitive analysis further shows that average vehicle delay decreases notably when the penetration rate of CAVs increases from 0% to 30%.

This article presents a whisper speech detector in the far-field domain. The proposed system consists of a long-short term memory (LSTM) neural network trained on log-filterbank energy (LFBE) acoustic features. This model is trained and evaluated on recordings of human interactions with voice-controlled, far-field devices in whisper and normal phonation modes. We compare multiple inference approaches for utterance-level classification by examining trajectories of the LSTM posteriors. In addition, we engineer a set of features based on the signal characteristics inherent to whisper speech, and evaluate their effectiveness in further separating whisper from normal speech. A benchmarking of these features using multilayer perceptrons (MLP) and LSTMs suggests that the proposed features, in combination with LFBE features, can help us further improve our classifiers. We prove that, with enough data, the LSTM model is indeed as capable of learning whisper characteristics from LFBE features alone compared to a simpler MLP model that uses both LFBE and features engineered for separating whisper and normal speech. In addition, we prove that the LSTM classifiers accuracy can be further improved with the incorporation of the proposed engineered features.

We present an end-of-utterance detector for real-time automatic speech recognition in far-field scenarios. The proposed system consists of three components: a long short-term memory (LSTM) neural network trained on acoustic features, an LSTM trained on l-best recognition hypotheses of the automatic speech recognition (ASR) decoder, and a feedforward deep neural network (DNN) combining embeddings derived from both LSTMs with pause duration features from the ASR decoder. At inference time, lower and upper latency (pause duration) bounds act as safeguards. Within the latency bounds, the utterance end-point is triggered as soon as the DNN posterior reaches a tuned threshold. Our experimental evaluation is carried out on real recordings of natural human interactions with voice-controlled far-field devices. We show that the acoustic embeddings are the single most powerful feature and particularly suitable for cross-lingual applications. We furthermore show the benefit of ASR decoder features, especially as a low cost alternative to ASR hypothesis em-beddings.

A method for speaker normalization in deep neural network (DNN) based discriminative feature estimation for automatic speech recognition (ASR) is presented. This method is applied in the context of a DNN configured for auto-encoder based low dimensional bottleneck (AE-BN) feature extraction where the derived features are used as input to a continuous Gaussian density hidden Markov model (HMM/GMM) based ASR decoder. While AE-BN features are known to provide significant reduction in ASR word error rate (WER) with respect to more conventional spectral magnitude based features, there is no general agreement on how these networks can reduce the impact of speaker variability by incorporating prior knowledge of the speaker. An approach is presented in this paper where spectrum based DNN inputs are augmented with speaker inputs that are derived from separate regression based speaker transformations. It is shown the proposed method could reduce the WER by 3% relative to the best speaker adapted AE-BN CDHMM system.

We present a comparative study on combination schemes for large vocabulary continuous speech recognition by incorporating long-span class posterior probability features into conventional short-time cepstral features. System combination can improve the overall speech recognition performance when multiple systems exhibit different error patterns and multiple knowledge sources encode complementary information. A variety of combination approaches are investigated in this paper, e.g., feature concatenation single stream system, model combination multi-stream system, lattice rescoring and ROVER. These techniques work at different levels of a LVCSR system and have different computational cost. We compared their performance and analyzed their advantages and disadvantages on large vocabulary English broadcast news transcription tasks. Experimental results showed that model combination with independent tree consistently outperforms ROVER, feature concatenation and lattice rescoring. In addition, the phoneme posterior probability features do provide complementary information to short-time cepstral features.

A detection-based paradigm decomposes a complex system into small pieces, solves each subproblem one by one, and combines the collected evidence to obtain a final solution. In this study of video story segmentation, a set of key events are first detected from heterogeneous multimedia signal sources, including a large scale concept ontology for images, text generated from automatic speech recognition systems, features extracted from audio track, and high-level video transcriptions. Then a discriminative evidence fusion scheme is investigated. We use the maximum figure-of-merit learning approach to directly optimize the performance metrics used in system evaluation, such as precision, recall, and F1 measure. Some experimental evaluations conducted on the TRECVID 2003 dataset demonstrate the effectiveness of the proposed detection-based paradigm. The proposed framework facilitates flexible combination and extensions of event detector design and evidence fusion to enable other related video applications.

In this paper, we propose an efficient gradient computation approach for discriminative fusion optimization in TRECVID high-level feature extraction. Numerical approximation was exploited in gradient calculation and model parameter update. The gradient of the performance measure was approximated by a sum of instance point-wise gradient instead of instance pair-wise gradient used in maximum figure-of-merit learning such that performance metrics like average precision can be optimized directly and efficiently on large training set. Experiments on the TRECVID 2005 high-level feature extraction test set showed that the proposed algorithm can improve the mean average precision from 0.254 of a state-of-the-art baseline system to 0.285.

This paper presents a novel unsupervised method for anchor shot detection using spectral clustering with multi-modal features. Unlike previous unsupervised studies where the acoustic trajectory features can not be combined with visual features directly, only a pairwise distance matrix from each attribute is needed instead of individual samples so that diverse information from heterogeneous features can be integrated in a unified manner. Experimental evaluation on a subset of the TRECVID 2004 dataset showed that an appropriate incorporation of the acoustic information with visual information will improve the F1 score from 0.68 for visual information only system to 0.87 in our unsupervised anchor shot detection system. Also a comparison study on the same dataset with a supervised system showed that the performance of our unsupervised system approach that of the supervised system.

In this paper, we address the task of identifying the speakers by name in audio content. Identification of speakers by name helps to improve the readability of the transcript and also provides additional meta-data which can help in finding the audio content of interest. We present a conditional maximum entropy (maxent) framework for this problem which yields superior performance and lends itself well to incorporating different types of information. We take advantage of this property of maxent to explore new features for this task. We show that supplementing standard lexical triggers with information such as speaker gender and position of speaker name mentions afford us large gains in performance. At 95% precision, we increase the recall to 67% from the trigger baseline of 38%.

We propose a new framework to incorporate speaker individuality parameters, such as pitch, vocal tract length and speaking rate, into designing speaker recognition systems. Based on our preliminary observations, a single pitch parameter may be more powerful than a vector of cepstral features for discriminating some speakers. Previous efforts have focused on concatenating these speaker parameters to existing MFCC based feature vectors. In this study a procedure is proposed to compare the effectiveness of the available set of parameters. The chosen parameter is then used to perform speaker verification. We test the proposed framework on the TIMIT database. Based on an intuitive parameter selection procedure to choose between a single pitch and the conventional 39-dim MFCC vector in a separate validation set, we found that parameter selection errors was reduced from 70, when only the MFCC parameter vector was used, to 25, when both parameter sets were made available in the selection process. For those 79 speakers whose corresponding pitch-based system was preferred for speaker verification, the average equal error rate was reduced from 23.1% to 18.4%. This strategy can be extended to incorporating other speaker individuality parameters

For efficient organization of speech recordings - meetings, interviews, voice mails, lectures - the ability to search for spoken keywords is an essential capability. Today, most spoken-document retrieval systems use large-vocabulary recognition. For the above scenarios, such systems suffer from both the unpredictable vocabulary/domain and generally high word-error rates (WER). We present a vocabulary-independent system to index and to search rapidly spontaneous speech. A speech recognizer generates lattices of phonetic word fragments, against which keywords are matched phonetically. We first show the need to use recognition alternatives (lattices) in a high-WER context, on a word-based baseline. Then we introduce our new method of phonetic word-fragment lattice generation, which uses longer-span language knowledge than a phoneme recognizer. Last we introduce heuristics to compact the lattices to feasible sizes that can be searched efficiently. On the LDC voice mail corpus, we show that vocabulary/domain-independent phonetic search is as accurate as a vocabulary/domain-dependent word-lattice based baseline system for in-vocabulary keywords (FOMs of 74-75%), but nearly maintains this accuracy also for out-of-vocabulary keywords.

The maximum likelihood estimation (MLE) and Bayesian maximum a-posteriori (MAP) adaptation methods for Gaussian mixture models (GMM) have proven to be effective and efficient for speaker verification, even though each speaker model is trained using only his own training utterances. Discriminative criteria aim at increasing discriminability by using out-of-class data. In this paper, we consider the speaker verification task using three discriminative training methods to compare performance. Comparisons are discussed for the maximum mutual information (MMI), minimum classification error (MCE) and figure of merit (FOM) criteria. Experiments on the 1996 NIST speaker recognition evaluation data set show that FOM training method outperforms the other two methods for speaker verification in terms of system performance. Meanwhile, logistic regression is investigated and successfully employed as a discriminative score-normalization technique.