One application for time-difference-of-arrival (TDOA) estimation is in emitter localization. A signal from an emitter reaching a group of sensors, each in a separate location, will have different arrival times. Finding the TDOAs between the output of pairs of sensors will provide the necessary measurements for the hyperbolic localization of the emitter. When the sensors acquire the signal by compressed sensing (CS), their outputs are reduced dimension linear transformation of the time samples of the signal. This shuffling of the time samples breaks up their time relation. Thus, a cross correlation of the CS output of two sensors cannot determine the TDOA. To apply cross correlation, it is necessary to reconstruct the time samples. This letter proposes an alternative that uses only the coefficients of the discrete Fourier transform (DFT) of the CS samples. It begins with the derivation of the maximum likelihood (ML) equation and the ML estimator. This estimator requires known values of signal and noise powers. Substituting these values by their estimates lead to the approximate ML estimator. The phase of the product of two DFT coefficients from each sensor is proportional to the unknown TDOA. Hence, these coefficients can provide an estimation of the TDOA. Simulation results show that although ML is the best, as expected, all these estimators have very close performance.

Emotional Internet of Things (EmIoT), which provides Internet of Things (IoT) devices cognitive and socialization capabilities, has been regarded as a future direction to improve users’ experiences. With the development of intelligent techniques, the requirement of EmIoT is not only sensing the users’ emotional states but also providing emotional feedbacks. Human–computer interaction has been studied to achieve speech interaction with IoT devices. The recent advances in neural text-to-speech (TTS) have made “human parity” synthesized speech possible for IoT-enabled human–computer interaction. Furthermore, emotion control can be achieved by using the emotional codes in a unified model, referred to as emotional TTS (or ETTS for short). Such ETTS models have achieved promising emotional expressiveness using large-scale emotion-annotated English data set; however, they are not practical in IoT environments with other mainstream languages, especially for Chinese. In fact, the limited available large-scale emotion-annotated data set is challenging the development of Chinese ETTS. To address that we propose a multistage deep transfer learning scheme to design a high-quality Chinese ETTS system under a small-scale training corpus to achieve EmIoT in Mandarin environments. In this scheme, the pretrained knowledge from the former stages corresponding to a large-scale neutral English and a medium-scale emotional English corpora is transferred to a Mandarin ETTS model. Thereby, the trained model can achieve high-quality emotional speech with limited available emotional corpus, which is able to serve various EmIoT-oriented applications. The experiments have been conducted to demonstrate the effectiveness and superiority of the proposed model as compared to other counterparts in terms of naturalness and emotional expressiveness. We refer readers to visit our demo Webpage ¹ enjoy the synthesized speech samples.

Estimation of the time-difference-of-arrival (TDOA) between a signal and its time-shifted version has many applications, mostly in localization. Because compressive sensing (CS) can reduce the sampling rate and data volume, TDOA estimation from CS samples (CS-TDOA) has become a subject of interest. Generally, CS-TDOA has lower estimation accuracy and the Cramer-Rao bound (CRB) can quantify this degradation. The first part of this correspondence derives the CRB for CSTDOA estimation. In the second part, it proposes an incoherent processing scheme for CS-TDOA estimation using partial Fourier coefficients. It has the advantage of reducing the number of sampling channels needed to acquire those coefficients. Simulation results are given to corroborate the theoretical development.

In this paper, random sampling is considered for direction-of-arrival (DOA) estimation with reduced hardware complexity in massive multiple-input–multiple-output (MIMO) system. The resulting problem is that the accuracy of the existing DOA estimators will significantly degrade with the availability of only a small subset of data entries as the low date rate is employed to reduce the system power consumption. To address that, an efficient approach based on the variant of matrix factorization is devised to complete the underlying data matrix. As a result, the nominal azimuth and elevation DOAs with the corresponding angular spreads are estimated from the underlying data matrix. Numerical results demonstrate that even in the case of missing entries, the proposed method is superior to the existing approaches with full data.

In this paper, direction-of-arrival (DOA) estimation in the presence of non-uniform noise is investigated. To achieve high estimation accuracy, the DOA estimation problem is solved with the use of a matrix completion approach. The matrix completion via alternating projection algorithm (APA) is firstly utilized to find the noise-free covariance matrix of the signal, which is then used to find the DOA by utilizing the conventional subspace-based method. Numerical simulations are performed to verify the accuracy and effectiveness of the proposed approach.

In this paper, a deep learning RNN model is used to classify Tibetan texts. The core idea is to first preprocess the Tibetan news corpus, and then use Tibetan syllables to construct a Tibetan syllable table based on the lexical and grammatical structure of Tibetan, embed the syllables in the sentence, and represent each syllable as a fixed Numerical vector. Secondly, the RNN cyclic neural network model is constructed. First, the text of different lengths is filled or truncated into a sequence length of uniform length. For each input text, the vector representation of text syllables is input on each time step of RNN to train the RNN model. The test samples were then used to evaluate the accuracy of model classification by introducing recall rate, precision rate and F-test. Finally, compared with traditional machine learning Logistic algorithm, polynomial naive Bayes algorithm and KNN algorithm, the results show that RNN model has better classification effect.