Automatic lexical stress and pitch accent detection for L2 English speech using multi-distribution deep neural networks

Li, Kun; Mao, Shaoguang; Xu, Li; Wu, Zhiyong; Meng, Helen

doi:10.1016/j.specom.2017.11.003

Cited by 51 publications

(33 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, each part has its own unique characteristics. In recent years, DNN algorithms have been a hot topic in the field of machine learning [25][26][27]. Among various detection tasks, it improves the recognition rate by a significant level.…”

Section: The Shared-weight Dnn Modulementioning

confidence: 99%

Anomaly Detection of CAN Bus Messages Using a Deep Neural Network for Autonomous Vehicles

Zhou

Shen

2019

Applied Sciences

View full text Add to dashboard Cite

The in-vehicle controller area network (CAN) bus is one of the essential components for autonomous vehicles, and its safety will be one of the greatest challenges in the field of intelligent vehicles in the future. In this paper, we propose a novel system that uses a deep neural network (DNN) to detect anomalous CAN bus messages. We treat anomaly detection as a cross-domain modelling problem, in which three CAN bus data packets as a group are directly imported into the DNN architecture for parallel training with shared weights. After that, three data packets are represented as three independent feature vectors, which corresponds to three different types of data sequences, namely anchor, positive and negative. The proposed DNN architecture is an embedded triplet loss network that optimizes the distance between the anchor example and the positive example, makes it smaller than the distance between the anchor example and the negative example, and realizes the similarity calculation of samples, which were originally used in face detection. Compared to traditional anomaly detection methods, the proposed method to learn the parameters with shared-weight could improve detection efficiency and detection accuracy. The whole detection system is composed of the front-end and the back-end, which correspond to deep network and triplet loss network, respectively, and are trainable in an end-to-end fashion. Experimental results demonstrate that the proposed technology can make real-time responses to anomalies and attacks to the CAN bus, and significantly improve the detection ratio. To the best of our knowledge, the proposed method is the first used for anomaly detection in the in-vehicle CAN bus.Appl. Sci. 2019, 9, 3174 2 of 12 start-stop, parking, Accessory system, and information entertainment systems that can be connected with smart devices such as mobile phones. These systems will obtain data from the CAN bus network on the vehicle. From the perspective of intelligent development, it is inevitable for automobiles to connect to the Internet, and these electronic devices and intelligent information systems may become a way for hackers to intrude into the automobile network system. Once hackers invade these systems and successfully connect to the car CAN bus network, the driver may lose control of the vehicle [3]. At the Black Hat conference in 2014, researchers released a report on the network security of more than 20 models on the market, assessing the ability of different car manufacturers to withstand malicious attacks on different models [4]. In addition, the (Identity Document) ID in the CAN bus protocol only represents the priority of the message, and there is no original address information in the protocol. The receiving electronic control unit (ECU) cannot confirm whether the received data is the original data or not; that is, the authenticity of the received message cannot be confirmed under the existing mechanism, which easily leads to forgery and tampering of the CAN bus message by injecting false information. Wha...

show abstract

Section: The Shared-weight Dnn Modulementioning

confidence: 99%

Anomaly Detection of CAN Bus Messages Using a Deep Neural Network for Autonomous Vehicles

Zhou

Shen

2019

Applied Sciences

View full text Add to dashboard Cite

show abstract

“…4, which is a multi-distribution DNN [15,16]. There are 273 Gaussian and 240 binary visible units in the bottom of the DNN.…”

Section: Structure Of Apmmentioning

confidence: 99%

“…The DNNs training in this work is similar to [14,16]. In the pre-training stage, we try to maximize the log-likelihood of RBMs.…”

Section: Dnns Trainingmentioning

confidence: 99%

See 1 more Smart Citation

Mispronunciation Detection and Diagnosis in L2 English Speech Using Multidistribution Deep Neural Networks

Qian

Meng

2017

IEEE/ACM Trans. Audio Speech Lang. Process.

Self Cite

123

View full text Add to dashboard Cite

This paper investigates the use of multi-distribution Deep Neural Networks (DNNs) for mispronunciation detection and diagnosis (MD&D). Our existing approach uses extended recognition networks (ERNs) to constrain the recognition paths to the canonical pronunciation of the target words and the likely phonetic mispronunciations. Although this approach is viable, it has some problems: (1) deriving appropriate phonological rules to generate the ERNs remains a challenging task; (2) the acoustic model (AM) and the phonological rules are trained independently and hence contextual information is lost; and (3) phones missing from the ERNs cannot be recognized even if we have a well-trained AM. Hence we propose an Acoustic Phonological Model (APM) using a multi-distribution DNN, whose input features include acoustic features and corresponding canonical pronunciations. The APM can implicitly learn the phonological rules from the canonical productions and annotated mispronunciations in the training data. Furthermore, the APM can also capture the relationships between the phonological rules and related acoustic features. As we do not restrict any pathways as in the ERNs, all phones can be recognized if we have a perfect APM. Experiments show that our method achieves an accuracy of 83.3% and a correctness of 88.5%. It significantly outperforms the approach of forced-alignment with ERNs whose correctness is 75.9%.

show abstract

“…An important aspect is also research in the field of voice analysis and processing. In [12] deep learning was used to quickly detect the accent for English language. Again in [13], the authors focused on recognizing emotions on the basis of voice samples using classical processing and classification techniques.…”

Section: Introductionmentioning

confidence: 99%

Voice control in mixed reality

Połap

2018

Annals of Computer Science and Information Systems

View full text Add to dashboard Cite

The gameplay in the augmented or virtual reality is based on the use of external equipment such as glasses/telephones and possibly the use of additional sensors or controllers. In both cases, the interaction involves pressing the keys on the phone or controller. An interesting aspect is the control of objects created in a virtual way using voice commands. In this paper, we propose a solution to manipulate objects in the augmented reality using player's voice. The user can move the object using pre-programmed commands. The solution is based on speech processing and artificial neural networks. The technique has been tested and the results presented and discussed.

show abstract

Automatic lexical stress and pitch accent detection for L2 English speech using multi-distribution deep neural networks

Cited by 51 publications

References 23 publications

Anomaly Detection of CAN Bus Messages Using a Deep Neural Network for Autonomous Vehicles

Anomaly Detection of CAN Bus Messages Using a Deep Neural Network for Autonomous Vehicles

Mispronunciation Detection and Diagnosis in L2 English Speech Using Multidistribution Deep Neural Networks

Voice control in mixed reality

Contact Info

Product

Resources

About