Low-latency System Research Articles

Expanding road noise cancellation to higher frequencies

Active vehicle Road Noise Cancellation (RNC) systems effectively reduce persistent vehicle rumble and booming noise which is very difficult to control with traditional NVH measures, particularly at low frequencies. Therefore, more and more vehicles are adapting RNC systems in their production applications. Most current RNC systems cancel low frequency structure-born noise, which is typically below 400 Hz. Although it provides appreciable noise attenuation, remaining high frequency noise contents still requires mitigation for a premium cabin experience. This study targets a next generation RNC system development which extends noise cancellation frequency range up to 1 kHz. High frequency RNC system (HF-RNC) development begins with analyzing limiting factors of the current RNC systems. Theoretical and practical limitations of the current system are discussed. Then, new technology and enablers to overcome those limitations are introduced. New enablers investigated herein in includes feedforward microphones, low latency system, headrest speakers and microphones, and head tracking system. HF-RNC system architecture is defined to comprise extended features and enablers. Low latency software is ported into a development platform which is based on a commercial DSP. The new HF-RNC system demonstrates noise cancellation up to 1 kHz when applied to an electric vehicle.

Low-latency remote-offloading system for accelerator

Specific workloads are increasingly offloaded to accelerators such as a graphic processing unit (GPU) and field-programmable gate array (FPGA) for real-time processing and computing efficiency. Because accelerators are expensive and consume much power, it is desirable to increase the efficiency of accelerator utilization by sharing accelerators among multiple servers over a network. However, task offloading over a network has the problem of latency due to network processing overhead in remote offloading. This paper proposes a low-latency system for accelerator offloading over a network. To reduce the overhead of remote offloading, we propose a system composed of (1) fast recombination processing of chunked data with a simple protocol to reduce the number of memory copies, (2) polling-based packet receiving check to reduce overhead due to interrupts in interaction with a network interface card, and (3) a run-to-completion model in network processing and accelerator offloading to reduce overhead with context switching. We show that the system can improve performance by 66.40% compared with a simple implementation using kernel protocol stack and confirmed the performance improvement with a virtual radio access network use case as a low-latency application. Furthermore, we show that this performance can also be achieved in practical usage in data center networks.

Causal speech enhancement using dynamical-weighted loss and attention encoder-decoder recurrent neural network.

Speech enhancement (SE) reduces background noise signals in target speech and is applied at the front end in various real-world applications, including robust ASRs and real-time processing in mobile phone communications. SE systems are commonly integrated into mobile phones to increase quality and intelligibility. As a result, a low-latency system is required to operate in real-world applications. On the other hand, these systems need efficient optimization. This research focuses on the single-microphone SE operating in real-time systems with better optimization. We propose a causal data-driven model that uses attention encoder-decoder long short-term memory (LSTM) to estimate the time-frequency mask from a noisy speech in order to make a clean speech for real-time applications that need low-latency causal processing. The encoder-decoder LSTM and a causal attention mechanism are used in the proposed model. Furthermore, a dynamical-weighted (DW) loss function is proposed to improve model learning by varying the weight loss values. Experiments demonstrated that the proposed model consistently improves voice quality, intelligibility, and noise suppression. In the causal processing mode, the LSTM-based estimated suppression time-frequency mask outperforms the baseline model for unseen noise types. The proposed SE improved the STOI by 2.64% (baseline LSTM-IRM), 6.6% (LSTM-KF), 4.18% (DeepXi-KF), and 3.58% (DeepResGRU-KF). In addition, we examine word error rates (WERs) using Google's Automatic Speech Recognition (ASR). The ASR results show that error rates decreased from 46.33% (noisy signals) to 13.11% (proposed) 15.73% (LSTM), and 14.97% (LSTM-KF).

Design and Build 4G Open Radio Access Network at SmartLab Politeknik Negeri Jakarta

The transition of RAN (Radio Access Network) development from the conventional model to Open RAN is motivated by the fact that Open RAN presents new services by reducing CAPEX (Capital Expenditure) and has the ability to integrate all cellular technologies in one server. One of cellular technology that can be developed using Open RAN is 4G LTE networks. The 4G LTE network produces internet with high data rate and low latency system. Various use case implementations can be carried out, for example to support learning on campus. In this study, the development of Open RAN 4G is implemented as a use case to build the Jakarta State Polytechnic SmartLab which produces its own cellular network. The importance of this research is to be able to ascertain the performance of the Open RAN 4G network using a Centralized RAN scenario. The method used in this research consists of knowing the problem, determining the tools used, determining the implementation scenario, configuring, and testing network performance. Parameters measured to test network performance are uplink downlink speed, RSRP value, and SINR value. Based on the scenarios that have been executed, the results show that the Open RAN 4G network has a maximum downlink speed performance of 13.6 Mbps, a maximum uplink speed of 7.22 Mbps, a maximum SINR of 22 dB, an RSRP of -200 dBm, and a maximum range that can be reached is 5 meters.

Simulated Hough Transform Model Optimized for Straight-Line Recognition Using Frontier FPGA Devices

The use of the Hough transforms to identify shapes or images has been extensively studied in the past using software for artificial intelligence applications. In this article, we present a generalization of the goal of shape recognition using the Hough transform, applied to a broader range of real problems. A software simulator was developed to generate input patterns (straight-lines) and test the ability of a generic low-latency system to identify these lines: first in a clean environment with no other inputs and then looking for the same lines as ambient background noise increases. In particular, the paper presents a study to optimize the implementation of the Hough transform algorithm in programmable digital devices, such as FPGAs. We investigated the ability of the Hough transform to discriminate straight-lines within a vast bundle of random lines, emulating a noisy environment. In more detail, the study follows an extensive investigation we recently conducted to recognize tracks of ionizing particles in high-energy physics. In this field, the lines can represent the trajectories of particles that must be immediately recognized as they are created in a particle detector. The main advantage of using FPGAs over any other component is their speed and low latency to investigate pattern recognition problems in a noisy environment. In fact, FPGAs guarantee a latency that increases linearly with the incoming data, while other solutions increase latency times more quickly. Furthermore, HT inherently adapts to incomplete input data sets, especially if resolutions are limited. Hence, an FPGA system that implements HT is inefficient for small sets of input data but becomes more cost-effective as the size of the input data increases. The document first presents an example that uses a large Accumulator consisting of 1100 × 600 Bins and several sets of input data to validate the Hough transform algorithm as random noise increases to 80% of input data. Then, a more specifically dedicated input set was chosen to emulate a real situation where a Xilinx UltraScale+ was to be used as the final target device. Thus, we have reduced the Accumulator to 280 × 280 Bins using a clock signal at 250 MHz and a few tens input points. Under these conditions, the behavior of the firmware matched the software simulations, confirming the feasibility of the HT implementation on FPGA.

Comparative Analysis of Feature Selection Techniques for Malicious Website Detection in SMOTE Balanced Data

The advancement in network technology has led to an exponential rise in the number of internet users across the globe. The increase in internet usage has resulted in an increase in both the number of malicious websites and cybercrimes reported over the years. Therefore, it has become critical to devise an intelligent solution that can detect malicious websites and be used in real-time systems. In our paper, we perform a comparative analysis of various feature selection techniques to build a time-efficient and accurate predictive model. To build our predictive model, a set of features are selected by feature selection methods. The selected features consist of at least 70% of the categorical features in all feature selection techniques examined in this paper. Keeping the end goal of real-time deployment of models in context the cost of processing or storing these features is far cheaper when compared to text or image-based features. We started out with a class imbalance in our data which was later dealt with using the Synthetic Minority Oversampling Technique. Our proposed model also bested the existing work in the literature when compared over various evaluation metrics. The result indicated that Embedded feature selection was the best technique considering the accuracy of the model. The Filter-based technique may also be used in the context of developing a low latency system at the cost of the accuracy of the model.

An efficient approach for low latency processing in stream data.

Stream data is the data that is generated continuously from the different data sources and ideally defined as the data that has no discrete beginning or end. Processing the stream data is a part of big data analytics that aims at querying the continuously arriving data and extracting meaningful information from the stream. Although earlier processing of such stream was using batch analytics, nowadays there are applications like the stock market, patient monitoring, and traffic analysis which can cause a drastic difference in processing, if the output is generated in levels of hours and minutes. The primary goal of any real-time stream processing system is to process the stream data as soon as it arrives. Correspondingly, analytics of the stream data also needs consideration of surrounding dependent data. For example, stock market analytics results are often useless if we do not consider their associated or dependent parameters which affect the result. In a real-world application, these dependent stream data usually arrive from the distributed environment. Hence, the stream processing system has to be designed, which can deal with the delay in the arrival of such data from distributed sources. We have designed the stream processing model which can deal with all the possible latency and provide an end-to-end low latency system. We have performed the stock market prediction by considering affecting parameters, such as USD, OIL Price, and Gold Price with an equal arrival rate. We have calculated the Normalized Root Mean Square Error (NRMSE) which simplifies the comparison among models with different scales. A comparative analysis of the experiment presented in the report shows a significant improvement in the result when considering the affecting parameters. In this work, we have used the statistical approach to forecast the probability of possible data latency arrives from distributed sources. Moreover, we have performed preprocessing of stream data to ensure at-least-once delivery semantics. In the direction towards providing low latency in processing, we have also implemented exactly-once processing semantics. Extensive experiments have been performed with varying sizes of the window and data arrival rate. We have concluded that system latency can be reduced when the window size is equal to the data arrival rate.

The CMS Trigger Upgrade for the HL-LHC

The CMS experiment has been designed with a two-level trigger system: the Level-1 Trigger, implemented on custom-designed electronics, and the High Level Trigger, a streamlined version of the CMS offline reconstruction software running on a computer farm. During its second phase the LHC will reach a luminosity of 7.5 1034 cm−2 s−1 with a pileup of 200 collisions, producing integrated luminosity greater than 3000 fb−1 over the full experimental run. To fully exploit the higher luminosity, the CMS experiment will introduce a more advanced Level-1 Trigger and increase the full readout rate from 100 kHz to 750 kHz. CMS is designing an efficient data-processing hardware trigger that will include tracking information and high-granularity calorimeter information. The current Level-1 conceptual design is expected to take full advantage of advances in FPGA and link technologies over the coming years, providing a high-performance, low-latency system for large throughput and sophisticated data correlation across diverse sources. The higher luminosity, event complexity and input rate present an unprecedented challenge to the High Level Trigger that aims to achieve a similar efficiency and rejection factor as today despite the higher pileup and more pure preselection. In this presentation we will discuss the ongoing studies and prospects for the online reconstruction and selection algorithms for the high-luminosity era.

Penjernihan Derau pada Suara Kanal Tunggal dengan Pembelajaran Faktorisasi Matriks Non-negatif tanpa Pengawasan

Artikel ini mengulas pendekatan metode penjernihan derau pada suara kanal tunggal menggunakan Faktorisasi Matriks Non-negatif (NMF) dengan pembelajaran tanpa pengawasan. Teknik ini memanfaatkan sifat NMF yang mengurai matriks spektrogram suara terganggu derau dan suara derau itu sendiri menjadi komponen vektor penyusunnya. Sebagai penunjang NMF, filter Wiener diterapkan pada akhir tahapan. Penjernihan ini digunakan untuk sistem dengan latensi rendah, sehingga menyediakan model derau secara khusus di awal proses secara terpisah menjadi tidak praktis. Maka dari itu model derau diambil langsung dari suara yang akan dijernihkan. Kontribusi yang dicapai dalam penelitian ini adalah jenis pembelajaran NMF dengan perbandingan konstrain linier dan non-linier yang dilakukan tanpa secara eksplisit menyediakan model derau, sehingga penjernihan dapat digunakan secara lebih fleksibel untuk setiap kondisi derau

Distance measurements and limitations based on guided wave 24 GHz dual tone Six-port radar

In the following, a continuous wave radar system based on the Six-port principle will be shown for measurement tasks at enclosed systems needing micrometer accuracy as well as high update rates like tank level monitoring or hydraulic cylinder piston control. To exceed the ambiguity limit of such an interferometric system, a dual tone approach is used. The system will be presented with measurement results at 24 GHz within a WR42 waveguide to prove the feasibility of the proposed concept. Furthermore, considerations on timing will show the potential as a low-latency system, capable of high measurement data update rates, and different influences on the system performance and limitations of such a system will be discussed in comparison to alternative setups.

Peer-to-Peer Live Multicast: A Video Perspective

Peer-to-peer multicast is promising for large-scale streaming video distribution over the Internet. Viewers contribute their resources to a peer-to-peer overlay network to act as relays for the media streams, and no dedicated infrastructure is required. As packets are transmitted over long, unreliable multipeer transmission paths, it is particularly challenging to achieve consistently high video quality and low end-to-end delay. In this paper, we focus on error-resilient transport for peer-to-peer video streaming. The algorithms we describe are representative of three broad categories of robust video streaming schemes: forward error correction, multiple descriptions, and prioritized automatic repeat request. We analyze how these techniques can be employed for live peer-to-peer multicast and discuss their relative merits. Our results show that significant gains can be obtained when systems are designed to adapt to the encoding structure of the video streams they are transmitting. They also reveal the importance of avoiding congestion at every peer participating in the multicast to obtain a low-latency system. Finally, we provide insights as to which are the important metrics to compare different peer-to-peer streaming systems.