Sound event detection is intended to identify the sound events in audio recordings, which has widespread applications in real life. Recently, convolutional recurrent neural network (CRNN) models have achieved state-of-the-art performance in this task due to their capabilities in learning the representative features. However, the CRNN models are of high complexities with millions of parameters to be trained, which limits their usage for the mobile and embedded devices with limited computation resource. Model distillation is effective to distill the knowledge of a complex model to a smaller one, which can be deployed on the devices with limited computational power. In this letter, we propose a novel multi model-based distillation approach for sound event detection by making use of the knowledge from models of multiple teachers which are complementary in detecting sound events. Extensive experimental results demonstrated that our approach achieves a compression ratio about 50 times. In addition, better performance is obtained for the sound event detection task.

This study designs a low-cost portable device that functions as a coaching assistant system which can support table tennis practice. Although deep learning technology is a promising solution to realizing human activity recognition, we propose using cosine similarity in making inferences. Our experiments show that the cosine similarity based inference can be a good alternative to the deep learning based inference for the assistant system when resources are limited.

A concurrent system consists of multiple processes that are run simultaneously. The execution orders of these processes are defined by a scheduler. In model checking techniques, the scheduling policy is closely related to the search algorithm that explores all of the system states. To ensure the correctness of the system, the scheduling policy needs to be taken into account during the verification. Current approaches, which use fixed strategies, are only capable of limited kinds of policies and are difficult to extend to handle the variations of the schedulers. To address these problems, we propose a method using a domain-specific language (DSL) for the succinct specification of different scheduling policies. Necessary artifacts are automatically generated from the specification to analyze the behaviors of the system. We also propose a search algorithm for exploring the state space. Based on this method, we develop a tool to verify the system with the scheduler. Our experiments show that we could serve the variations of the schedulers easily and verify the systems accurately.

Sharing perceptual data (e.g., camera and LiDAR data) with other vehicles enhances the traffic safety of autonomous vehicles because it helps vehicles locate other vehicles and pedestrians in their blind spots. Such safety applications require high throughput and short delay, which cannot be achieved by conventional microwave vehicular communication systems. Therefore, millimeter-wave (mmWave) communications are considered to be a key technology for sharing perceptual data because of their wide bandwidth. One of the challenges of data sharing in mmWave communications is broadcasting because narrow-beam directional antennas are used to obtain high gain. Because many vehicles should share their perceptual data to others within a short time frame in order to enlarge the areas that can be perceived based on shared perceptual data, an efficient scheduling for concurrent transmission that improves spatial reuse is required for perceptual data sharing. This paper proposes a data sharing algorithm that employs a graph-based concurrent transmission scheduling. The proposed algorithm realizes concurrent transmission to improve spatial reuse by designing a rule that is utilized to determine if the two pairs of transmitters and receivers interfere with each other by considering the radio propagation characteristics of narrow-beam antennas. A prioritization method that considers the geographical information in perceptual data is also designed to enlarge perceivable areas in situations where data sharing time is limited and not all data can be shared. Simulation results demonstrate that the proposed algorithm doubles the area of the cooperatively perceivable region compared with a conventional algorithm that does not consider mmWave communications because the proposed algorithm achieves high-throughput transmission by improving spatial reuse. The prioritization also enlarges the perceivable region by a maximum of 20%.

To solve the low accuracy problem of the recommender system for long term users, in this paper, we propose a top-N-balanced sequential recommendation based on recurrent neural network. We postulated and verified that the interactions between users and items is time-dependent in the long term, but in the short term, it is time-independent. We balance the top-N recommendation and sequential recommendation to generate a better recommender list by improving the loss function and generation method. The experimental results demonstrate the effectiveness of our method. Compared with a state-of-the-art recommender algorithm, our method clearly improves the performance of the recommendation on hit rate. Besides the improvement of the basic performance, our method can also handle the cold start problem and supply new users with the same quality of service as the old users.

Business process similarity measure is required by many applications, such as business process query, improvement, redesign, and etc. Many process behavior similarity measures have been proposed in the past two decades. However, to the best of our knowledge, most existing work only focuses on the direct causality transition relations and totally neglect the concurrent and transitive transition relations that are proved to be equally important when measuring process behavior similarity. In this paper, we take the weakness of existing process behavior similarity measures as a starting point, and propose a comprehensive approach to measure the business process behavior similarity based on the so-called Extended Transition Relation set, ETR-set for short. Essentially, the ETR-set is an ex-tended transition relation set containing direct causal transition relations, minimum concurrent transition relations and transitive causal transition relations. Based on the ETR-set, a novel process behavior similarity measure is defined. By constructing a concurrent reachability graph, our approach finds an effective technique to obtain the ETR-set. Finally, we evaluate our proposed approach in terms of its property analysis as well as conducting a group of control experiments.

Coil-shaped structures are proposed to enhance sensitivity and spatial resolution for EMI near-field probe. This design yields a high sensitivity and a good spatial resolution to find the EMI source in near-field region. Both characteristics are crucial to diagnosis of emissions from electrical and electronic devices. The new design yields a superior sensitivity, which is in general 15 dB greater than conventional probes. This new probe helps practitioners to quickly and correctly locate noise emission source areas on printed circuit boards and devices. Two prototypes of different sizes were fabricated. The larger one provides a high sensitivity while the smaller one can pinpoint emission source locations. The new probe design also has an orientation invariance feature. Its noise response levels are similar for all probe directions. This characteristic can help reduced the probability at miss-detection since sensitivity is largely invariant to its orientation. Extensive measurements were performed to verify the operation mechanism and to assess probe characteristics. It suits well to the electromagnetic interference problem diagnosis.

On surfaces of tris-(8-hydroxyquinolate) aluminum (Alq) and tris(7-propyl-8-hydroxyquinolinato) aluminum (Al7p) thin-films, positive and negative polarization charges appear, respectively, owing to spontaneous orientation of these polar molecules. Alq is a typical electron transport material where electrons are injected from cathode. Because the polarization charge exists at the Alq/cathode interface, it is likely that it affects the electron injection process because of Coulomb interaction. In order to evaluate an impact of polarization charge on electron injection from cathode, electron only devices (EODs) composed of Alq or Al7p were prepared and evaluated by displacement current measurement. We found that Alq-EOD has lower resistance than Al7p-EOD, indicating that the positive polarization charge at Alq/cathode interface enhances the electron injection due to Coulomb attraction, while the electron injection is suppressed by the negative polarization charge at the Al7p/Al interface. These results clearly suggest that it is necessary to design organic semiconductor devices by taking polarization charge into account.

Two kinds of composite right/left-handed coaxial lines (CRLH CLs) are designed for an antenna element. The dispersion relations of the infinite periodic CRLH CLs are designed to occur -1st resonance at around 700 MHz, respectively. The designed CRLH CLs comprise a monopole and a choke structure for antenna elements. To verify the resonant modes and frequencies, the monopole structure, the choke structure, and the antenna element which is combined the monopole and the choke structures are simulated by eigenmode analysis. The resonant frequencies correspond to the dispersion relations. The monopole and the choke structures are applied to the coaxially fed antenna. The proposed antenna matches at 710 MHz and radiates. At the resonant frequency, the total length of the proposed antenna which is the length of the monopole structure plus the choke structure is 0.12 wavelength. The characteristics of the proposed antenna has been compared with that of the conventional coaxially fed monopole antenna without the choke structure and the sleeve antenna with the quarter-wavelength choke structure. The radiation pattern of the proposed antenna is omnidirectional, the total antenna efficiency is 0.73 at resonant frequencies, and leakage current is suppressed lesser than -10 dB at resonant frequency. The propose antenna is fabricated and measured. The measured |S11| characteristics, radiation patterns, and the total antenna efficiency are in good agreement with the simulated results.

In this study, transverse magnetic electromagnetic plane wave scatterings by rectangular cylinders have been analyzed by a high frequency asymptotic method. Scattering field can be generated by the equivalent electric and magnetic currents which are obtained approximately from the geometrical optics (GO) fields. Our formulation is found to be exactly the same with the physical optics (PO) for the conducting cylinders, and it can also be applicable for dielectric cylinders. Numerical calculations are made to compare the results with those by other methods, such as the geometrical theory of diffraction (GTD) and HFSS simulation. A good agreement has been observed to confirm the validity of our method.

Forward converter is most suitable for low voltage and high current applications such as LEDs, battery chargers, EHV etc. The active clamp transformer reset technique offers many advantages over conventional single-ended reset techniques, including lower voltage stress on the main switch, the ability to switch at zero voltage and duty cycle operation above 50 percent. Several papers have compared the functional merits of the active clamp over the more extensively used RCD clamp, third winding and resonant reset techniques. This paper discusses about a center clamp technique with one common core reset circuit making it suitable for wide input voltage applications with extended duty cycle.

This study proposes a multiple-output differential-input operational transconductance amplifier-C (MODI OTA-C) filter with an impedance scaler to detect cardiac activity. A ladder-type fifth-orderButterworth low-pass filter with a large time constant and low noise is implemented to reduce coefficient sensitivity and address signal distortion. Moreover, linearized MODI OTA structures with reduced transconductance and impedance scaler circuits for noise reduction are used to achieve a wide dynamic range (DR). The OTA-based circuit is operated in the subthreshold region at a supply voltage of 1 V to reduce the power consumption of the wearable device in long-term use. Experimental results of the filter with a bandwidth of 250 Hz reveal that DR is 57.6 dB, and the harmonic distortion components are below -59 dB. The power consumption of the filter, which is fabricated through a TSMC 0.18 µm CMOS process, is lower than 390 nW, and the active area is 0.135 mm2.

A fusion framework between CNN and RNN is proposed dedicatedly for air-writing recognition. By modeling the air-writing using both spatial and temporal features, the proposed network can learn more information than existing techniques. Performance of the proposed network is evaluated by using the alphabet and numeric datasets in the public database namely the 6DMG. Average accuracy of the proposed fusion network outperforms other techniques, i.e. 99.25% and 99.83% are observed in the alphabet gesture and the numeric gesture, respectively. Simplified structure of RNN is also proposed, which can attain about two folds speed-up of ordinary BLSTM network. It is also confirmed that only the distance between consecutive sampling points is enough to attain high recognition performance.

We developed and applied a new circuit, called the “Self-controllable Voltage Level (SVL)” circuit, not only to expand both “write” and “read” stabilities, but also to achieve a low stand-by power and data holding capability in a single low power supply, 90-nm, 2-kbit, six-transistor CMOS SRAM. The SVL circuit can adaptively lower and higher the word-line voltages for a “read” and “write” operation, respectively. It can also adaptively lower and higher the memory cell supply voltages for the “write” and “hold” operations, and “read” operation, respectively. This paper focuses on the “hold” characteristics and the standby power dissipations (PST) of the developed SRAM. The average PST of the developed SRAM is only 0.984µW, namely, 9.57% of that (10.28µW) of the conventional SRAM at a supply voltage (VDD) of 1.0V. The data hold margin of the developed SRAM is 0.1839V and that of the conventional SRAM is 0.343V at the supply voltage of 1.0V. An area overhead of the SVL circuit is only 1.383% of the conventional SRAM.

The-state-of-the-art neural quality estimation (QE) of machine translation model consists of two sub-networks that are tuned separately, a bidirectional recurrent neural network (RNN) encoder-decoder trained for neural machine translation, called the predictor, and an RNN trained for sentence-level QE tasks, called the estimator. We propose to combine the two sub-networks into a whole neural network, called the unified neural network. When training, the bidirectional RNN encoder-decoder are initialized and pre-trained with the bilingual parallel corpus, and then, the networks are trained jointly to minimize the mean absolute error over the QE training samples. Compared with the predictor and estimator approach, the use of a unified neural network helps to train the parameters of the neural networks that are more suitable for the QE task. Experimental results on the benchmark data set of the WMT17 sentence-level QE shared task show that the proposed unified neural network approach consistently outperforms the predictor and estimator approach and significantly outperforms the other baseline QE approaches.

To learn the working situation of rod-pumped wells under ground, we always need to analyze dynamometer diagrams, which are generated by the load sensor and displacement sensor. Rod-pumped wells are usually located in the places with extreme weather, and these sensors are installed on some special oil equipments in the open air. As time goes by, sensors are prone to generating unstable and incorrect data. Unfortunately, load sensors are too expensive to frequently reinstall. Therefore, the resulting dynamometer diagrams sometimes cannot make an accurate diagnosis. Instead, as an absolutely necessary equipment of the rod-pumped well, the electric motor has much longer life and cannot be easily impacted by the weather. The electric power curve during a swabbing period can also reflect the working situation under ground, but is much harder to explain than the dynamometer diagram. This letter presented a novel deep learning architecture, which can transform the electric power curve into the dimensionless dynamometer diagram image. We conduct our experiments on a real-world dataset, and the results show that our method can get an impressive transformation accuracy.

The traditional Markov prediction methods of the taxi destination rely only on the previous 2 to 3 GPS points. They negelect long-term dependencies within a taxi trajectory. We adopt a Recurrent Neural Network (RNN) to explore the long-term dependencies to predict the taxi destination as the multiple hidden layers of RNN can store these dependencies. However, the hidden layers of RNN are very sensitive to small perturbations to reduce the prediction accuracy when the amount of taxi trajectories is increasing. In order to improve the prediction accuracy of taxi destination and reduce the training time, we embed suprisal-driven zoneout (SDZ) to RNN, hence a taxi destination prediction method by regularized RNN with SDZ (TDPRS). SDZ can not only improve the robustness of TDPRS, but also reduce the training time by adopting partial update of parameters instead of a full update. Experiments with a Porto taxi trajectory data show that TDPRS improves the prediction accuracy by 12% compared to RNN prediction method in literature[4]. At the same time, the prediction time is reduced by 7%.

The 1-bit band-pass delta-sigma modulator (BP-DSM) achieves high resolution if it uses an oversampling technique. This method can generate concurrent dual-band RF signals from a digitally modulated signal using a 1-bit digital pulse train. It was previously reported that the adjacent channel leakage ratio (ACLR) deteriorates owing to the asymmetrical waveform created by the pulse transition mismatch error of the rising and falling waveforms in the time domain and that the ACLR can be improved by distortion compensation. However, the reported distortion compensation method can only be performed for single-band transmission, and it fails to support multi-band transmission because the asymmetrical waveform compensated signal extends over a wide frequency range and is itself a harmful distortion outside the target band. Unfortunately, the increase of out-of-band power causes the BP-DSM unstable. We therefore propose a distortion compensator for a concurrent dual-band 1-bit BP-DSM that consists of a noise transfer function with a quasi-elliptic filter that can control the out-of-band gain frequency response against out-of-band oscillation. We demonstrate that dual-band LTE signals, each with 40MHz (2×20MHz) bandwidth, at 1.5 and 3.0GHz, can be compensated concurrently for spurious distortion under various combinations of rising and falling times and ACLR of up to 48dB, each with 120MHz bandwidth, including the double sided adjacent channels and next adjacent channels, is achieved.

Automatic speech recognition (ASR) and keyword search (KWS) have more and more found their way into our everyday lives, and their successes could boil down lots of factors. In these factors, large scale of speech data used for acoustic modeling is the key factor. However, it is difficult and time-consuming to acquire large scale of transcribed speech data for some languages, especially for low-resource languages. Thus, at low-resource condition, it becomes important with which transcribed data for acoustic modeling for improving the performance of ASR and KWS. In view of using acoustic data for acoustic modeling, there are two different ways. One is using the target language data, and another is using large scale of other source languages data for cross-lingual transfer. In this paper, we propose some approaches for efficient selecting acoustic data for acoustic modeling. For target language data, a submodular based unsupervised data selection approach is proposed. The submodular based unsupervised data selection could select more informative and representative utterances for manual transcription for acoustic modeling. For other source languages data, the high misclassified as target language based submodular multilingual data selection approach and knowledge based group multilingual data selection approach are proposed. When using selected multilingual data for multilingual deep neural network training for cross-lingual transfer, it could improve the performance of ASR and KWS of target language. When comparing our proposed multilingual data selection approach with language identification based multilingual data selection approach, our proposed approach also obtains better effect. In this paper, we also analyze and compare the language factor and the acoustic factor influence on the performance of ASR and KWS. The influence of different scale of target language data on the performance of ASR and KWS at mono-lingual condition and cross-lingual condition are also compared and analyzed, and some significant conclusions can be concluded.

In this paper, we propose the novel low voltage CMOS current mode reference circuit. It reduces the minimum supply voltage by consisting the subthreshold two stage operational amplifier (OPAMP) which is regarded as the combination of the proportional to absolute temperature (PTAT) and the complementary to absolute temperature (CTAT) current generators. It makes possible to implement without extra OPAMP. This proposed circuit has been designed and evaluated by SPICE simulation using TSMC 65nm CMOS process with 3.3V (2.5V over-drive) transistor option. From simulation results, the line sensitivity is as good as 0.196%/V under the condition that the range of supply voltage (VDD) is wide as 0.6V to 3.0V. The temperature coefficient is 71ppm/ under the condition that the temperature range is from -40 to 125 and VDD=0.6V. The power supply rejection ratio (PSRR) is -47.7dB when VDD=0.6V and the noise frequency is 100Hz. According to comparing the proposed circuit with prior current mode circuits, we could confirm the performance of the proposed circuit is better than that of prior circuits.