A High-throughput Parallel Viterbi Algorithm via Bitslicing

In this work, we present a novel bitsliced high-performance Viterbi algorithm suitable for high-throughput and data-intensive communication. A new column-major data representation scheme coupled with the bitsliced architecture is employed in our proposed Viterbi decoder that enables the maximum utilization of the parallel processing units in modern parallel accelerators. With the help of the proposed alteration of the data scheme, instead of the conventional bit-by-bit operations, 32-bit chunks of data are processed by each processing unit. This means that a single bitsliced parallel Viterbi decoder is capable of decoding 32 different chunks of data simultaneously. Here, the Viterbi’s Add-Compare-Select procedure is implemented with our proposed bitslicing technique, where it is shown that the bitsliced operations for the Viterbi internal functionalities are efficient in terms of their performance and complexity. We have achieved this level of high parallelism while keeping an acceptable bit error rate performance for our proposed methodology. Our suggested hard and soft-decision Viterbi decoder implementations on GPU platforms outperform the fastest previously proposed works by 4.3{\times } and 2.3{\times } , achieving 21.41 and 8.24 Gbps on Tesla V100, respectively.

Data-driven model for ship encounter intention inference in intersection waters based on AIS data

Frequent collision accidents of ships in intersection waters have caused huge casualties and property losses. Unclear encounter intention, poor communication, or inaccurate judgment of the encounter intention are often the major causes of ships falling into dangerous and urgent situations, leading to collision accidents. There are few methods and models for automatically inferring ship encounter intention. In this study, an intelligent model driven by AIS data is proposed to infer the ship encounter intention in intersection waters. The Hidden Markov Model (HMM) is adopted to formulate the encounter process and perform intention inference. The encounter intentions, including crossing, overtaking and head-on, are modeled as unobservable states of the formulated HMM. The observable measures of HMM extracted from AIS data, include the relative distance, relative speed, and course difference between two ships. Subsequently, the Forward-Backward algorithm is employed to obtain the model parameters and the Viterbi algorithm is exploited to estimate the hidden state with the highest probability, resulting in the inferred intention. The main advantage of the proposed model is its ability to capture the spatial-temporal characteristics of the encounter process, that is, the spatial interaction between ships and the dynamic evolution of states of the encounter process. The AIS data collected from the Lantau Strait intersection waters are adopted to verify the effectiveness of the proposed model. The experimental results reveal that the model can achieve an inference accuracy of 95%, 91.33%, and 92.67% for crossing, overtaking, and head-on, respectively. Moreover, it has real-time performance that ensures the encounter intentions can be recognized at an early stage, which is very critical for the safe navigation of any ships encountered. Our results show that our model can infer the encounter intentions in a timely manner and with high accuracy.

Design of ACS Architecture Using FinFET and CNTFET Devices for Low-Power Viterbi Decoder Using Asynchronous Techniques for Digital Communication Systems

Viterbi algorithm is the most popular algorithm used to decode the convolution code, but its computational complexity increases exponentially with the increasing constraint length due to a large number of Trellis transitions. However, high constraint length is necessary to improve the accuracy of the decoding process for the high rate convolution code. In particular, the Add-Compare-Select (ACS) module of the Viterbi Decoder will have large numbers of trellis states and trellis transitions with increased constraint lengths, which give rise to high hardware complexity and large power consumption. As the performance of the Viterbi decoder mainly depends on its efficient implementation of the ACS module, in the literature, several methods are presented for the implementation of ACS for the Viterbi decoder. The methods based on Precharge Half Buffer (PCHB) and Weak Conditioned Half Buffer, Shannon’s decomposition circuits, body-biased pseudo-NMOS logic and Quasi Delay Insensitive (QDI) timing model performance is analyzed. The methods are implemented using CMOS technology. In this paper, FinFET and CNTFET-based ACS implementation is performed. From the analysis, it has been found that the Carbon Nanotube-based implementation is better in performance when compared to the CMOS and FinFET technology. The proposed QDI model and retiming circuits for ACS block operate above 1[Formula: see text]GHz with high driving current and low power.

NEW CONCATENATION SCHEMES BASED ON THE MULTITHRESHOLD DECODERS OF CONVOLUTIONAL SELF-ORTHOGONAL CODES FOR GAUSSIAN CHANNELS

New serial concatenation schemes based on the multithreshold decoders and di- vergent principle for the convolutional self-orthogonal codes under Gaussian channels are proposed. Using both binary and symbolic decoders on the second decoding stage of the convolutional codes are considered. Simulation results are indicated the higher performance characteristics of the proposed cascade schemes on majority decoders in comparison with clas- sical schemes based on the Viterbi algorithm and Reed-Solomon codes. A moderate increase in decoding delay during concatenation is revealed. It is determined by the absence of the need to use traditional two-dimensional concatenated structures.

Klasifikasi Kelas Kata (Part-Of-Speech Tagging) untuk Bahasa Madura Menggunakan Algoritme Viterbi

<p class="Abstrak">Bahasa manusia adalah bahasa yang digunakan oleh manusia dalam bentuk tulisan maupun suara. Banyak teknologi/aplikasi yang mengolah bahasa manusia, bidang tersebut bernama <em>Natural Language Processing </em>yang merupakan ilmu yang mempelajari untuk mengolah dan mengekstraksi bahasa manusia pada perkembangan teknologi. Salah satu proses pada <em>Natural Language Processing </em>adalah <em>Part-Of-Speech Tagging</em>. <em>Part-Of-Speech Tagging </em>adalah klasifikasi kelas kata pada sebuah kalimat secara otomatis oleh teknologi, proses ini salah satunya berfungsi untuk mengetahui kata-kata yang memiliki lebih dari satu makna/arti (ambiguitas). <em>Part-Of-Speech Tagging</em> merupakan dasar dari <em>Natural Language Processing</em> lainnya, seperti penerjemahan mesin (<em>machine translation</em>), penghilangan ambiguitas makna kata (<em>word sense disambiguation</em>), dan analisis sentimen. <em>Part-Of-Speech Tagging</em> dilakukan pada bahasa manusia, salah satunya adalah bahasa Madura. Bahasa Madura adalah bahasa daerah yang digunakan oleh suku Madura dan memiliki morfologi yang mirip dengan bahasa Indonesia. Penelitian pada <em>Part-Of-Speech Tagging </em>pada bahasa Madura ini menggunakan algoritme Viterbi, terdapat 3 proses untuk implementasi algoritme Viterbi pada pada <em>Part-Of-Speech Tagging</em> bahasa Madura, yaitu <em>pre-processing </em>pada data<em> training </em>dan <em>testing</em>, perhitungan data latih dengan <em>Hidden Markov Model </em>dan klasifikasi kelas kata menggunakan algoritme Viterbi. Kelas kata (<em>tagset</em>) yang digunakan untuk klasifikasi kata pada bahasa Madura sebanyak 19 kelas, kelas kata tersebut dirancang oleh pakar. Pengujian sistem pada penelitian ini menggunakan perhitungan <em>Multiclass Confusion Matrix</em>. Hasil pengujian sistem mendapatkan nilai <em>micro average</em> <em>accuracy </em>sebesar 0,96 dan nilai <em>micro average</em> <em>precision </em>dan <em>recall </em>yang sama sebesar 0,68. <em>Precision</em> dan <em>recall</em> masih dapat ditingkatkan dengan menambahkan data yang lebih banyak lagi untuk pelatihan.</p><p class="Abstrak"> </p><p class="Abstrak"><em><strong>Abstract</strong></em></p><p class="Abstract"><em>Natural language is a form of language used by human, either in writing or speaking form. There is a specific field in computer science that processes natural language, which is called Natural Language Processing. It is a study of how to process and extract natural language on technology development. Part-Of-Speech Tagging is a method to assign a predefined set of tags (word classes) into a word or a phrase. This process is useful to understand the true meaning of a word with ambiguous meaning, which may have different meanings depending on the context. Part-Of-Speech Tagging is the basis of the other Natural Language Processing methods, such as machine translation, word sense disambiguation, and sentiment analysis. Part-Of-Speech Tagging used in natural languages, such as Madurese language. Madurese language is a local language used by Madurese and has a similar morphology as Indonesian language. Part-Of-Speech Tagging research on Madurese language using Viterbi algorithm, consists of 3 processes, which are training and testing corpus pre-processing, training the corpus by Hidden Markov Model, and tag classification using Viterbi algorithm. The number of tags used for words classification (tagsets) on Madurese language are 19 class, those tags were designed by an expert. Performance assessment was conducted using Multiclass Confusion Matrix calculation. The system achieved a micro average accuracy score of 0,96, and micro average precision score is equal to recall of 0,68. Precision and recall can still be improved by adding more data for training.</em></p><p class="Abstrak"><em><strong><br /></strong></em></p>

An Optimization Approach to Multi-Sensor Operation for Multi-Context Recognition

Mobile devices and sensors have limited battery lifespans, limiting their feasibility for context recognition applications. As a result, there is a need to provide mechanisms for energy-efficient operation of sensors in settings where multiple contexts are monitored simultaneously. Past methods for efficient sensing operation have been hierarchical by first selecting the sensors with the least energy consumption, and then devising individual sensing schedules that trade-off energy and delays. The main limitation of the hierarchical approach is that it does not consider the combined impact of sensor scheduling and sensor selection. We aimed at addressing this limitation by considering the problem holistically and devising an optimization formulation that can simultaneously select the group of sensors while also considering the impact of their triggering schedule. The optimization solution is framed as a Viterbi algorithm that includes mathematical representations for multi-sensor reward functions and modeling of user behavior. Experiment results showed an average improvement of 31% compared to a hierarchical approach.

Analisis Kinerja FBMC OQAM menggunakan Kode Konvolusi

ABSTRAKSistem FBMC merupakan teknologi MCM yang dapat menyediakan laju data bit yang tinggi. Modulasi digital OQAM digunakan untuk meningkatkan bit rate. Pengkodean kanal digunakan untuk mengoreksi kesalahan yang diakibatkan noise. Penilitian ini menggunakan pengkodean kanal kode konvolusi yang digunakan pada bagian pengirim dan algortima viterbi pada bagian penerima. Simulasi dilakukan pada FBMC OQAM dengan kode konvolusi dan tanpa kode konvolusi dengan perbandingan parameter BER dan kapasitas kanal terhadap SNR. Hasil penelitian menunjukan FBMC OQAM dengan kode konvolusi lebih baik daripada FBMC OQAM tanpa kode konvolusi pada SNR tinggi. Pada FBMC OQAM untuk mencapai BER 10-3 membutuhkan SNR 17 dB sedangkan pada FBMC OQAM dengan kode konvolusi membutuhkan SNR 16 dB. Peningkatan SNR dapat meningkatkan kapasitas kanal yang dihasilkan, pada SNR 0 dB menghasilkan 0,4535 bps/Hz dan SNR 20 dB menghasilkan 5,858 bps/Hz.Kata kunci: kode konvolusi, algoritma viterbi, FBMC, OQAM, BER ABSTRACTThe FBMC system is an MCM technology that can provide high bit data rates. OQAM digital modulation is used to increase the bit rate. Channel coding is used to correct errors caused by noise. This research uses convolutional code channel coding used on the sender and viterbi algorithms on the receiver. Simulations are carried out on FBMC OQAM with convolutional code and without convolutional code with a comparison of BER parameters and channel capacity to SNR. The results showed that FBMC OQAM with convolutional code was better than FBMC OQAM without convolutional code at high SNR. In FBMC OQAM to reach BER 10-3 requires SNR of 17 dB while in FBMC OQAM with convolutional code requires SNR of 16 dB. Increasing SNR can increase the resulting channel capacity, at 0 dB SNR it produces 0.4535 bps / Hz and SNR 20 dB produces 5.858 bps / Hz.Keywords: convolutional code, viterbi algorithm, FBMC, OQAM, BER

External knowledge transfer deployment inside a simple double agent Viterbi algorithm

We consider in this paper deploying external knowledge transfer inside a simple double agent Viterbi algorithm which is an algorithm firstly introduced by the author in his preprint "Hidden Markov Based Mathematical Model dedicated to Extract Ingredients from Recipe Text". The key challenge of this work lies in discovering the reason why our old model does have bad performances when it is confronted with estimating ingredient state for unknown words and see if deploying external knowledge transfer directly on calculating state matrix could be the solution instead of deploying it only on back propagating step.