KJKamran Javed

Kamran Javed

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Kamran Javed has a total of 17 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Mitigating Common Mode Current in Chain Energy Storage Inverters With Novel Approach

    by Qinglei Zhang, Waseem Ahmad, Kamran Javed, Ali Raza

    The chain energy storage converter, a type of multilevel converter, exhibits characteristics such as multi-modularization, easy scalability, and low harmonic voltage output, rendering it suitable for medium to high voltage and large capacity energy storage systems. However, the large volume of battery cabinets in large-scale energy storage systems introduces parasitic parameters into the cascaded H-bridge converter, leading to significant common-mode current during switching operations. This paper presents a novel suppression scheme for the chain energy storage inverter's common-mode current. The proposed approach incorporates symmetrical filter inductors and one-dimensional vector modulation to minimize common-mode current and voltage. Simulations conducted in MATLAB/Simulink confirm the effectiveness of the proposed strategy, showcasing substantial reductions in common-mode current and improved compliance with grid specifications.

  • Open Access

    Energy-Efficient Cell Association and Load Balancing for Low Battery Users in Heterogeneous Cellular Networks

    by Abdul Qadeer Khan, Kamran Javed, Ali Raza, Shahryar Saleem, Zubair Saeed

    The future Heterogeneous Cellular Network (HCN) is expected to fulfill the demand for the Internet of Things, Industry 4.0, extended reality, connected vehicles and mobile gadgets. In HCN, the deployment of low power nodes inside the coverage area of a high power node yields the benefit of providing improved data rates to users by offloading a portion of users from a highly loaded high power node to relatively lightly loaded low power nodes. However, the offloading of users through conventional cell association schemes may not guarantee improved load balancing along with the energy efficacy (EE) of critical users such as user equipment (UE) with low battery power. This paper proposes improved cell association schemes based on the battery levels of UE and Deep Q-learning (DQL) to achieve load balancing and to decrease the power consumption of Low Battery Users (LBUs). The proposed approach combines UE battery levels with three conventional association schemes for improvement: minimum distance based cell association, maximum biased received power (MBRP), and maximum biased SINR (MBSINR). Results on cell association, load balancing, and EE of LBUs have been compared with conventional association schemes. The proposed battery based cell association models perform well in favor of LBUs and clearly show the significance of the proposed approach.

  • Open Access

    Code-Mixed Street Address Recognition and Accent Adaptation for Voice-Activated Navigation Services

    by Syed Meesam Raza Naqvi, Muhammad Ali Tahir, Kamran Javed, Hassan Aqeel Khan, Ali Raza, Zubair Saeed

    This study presents the development of a real-time application-specific Automatic Speech Recognition (ASR) system for voice-activated navigation services. The system is designed to recognize Urdu-English code-mixed street addresses, which is challenging due to their complex nature and structure, especially in under-resourced languages such as Urdu. Two separate corpora are collected for ASR system development: Unicode Urdu consisting of general Urdu recordings of around 61.82 hours by 144 speakers and Roman Urdu-English code-mixed Addresses of around 16.89 hours by 20 speakers. The Unicode Urdu data is developed to provide acoustic models with general language understanding and code-mixed street addresses to provide code-mixing or switching coverage. The hybrid ASR system employed in this study plays a crucial role in addressing the multifaceted challenges of low-resource settings (only 16.89 hours of task-specific data), especially in the context of Urdu-English code-switching. The study compares various acoustic models, with mixed Time Delay Neural Network and Long Short-Term Memory (TDNN-LSTM) performing best with a Word Error Rate (WER), Character Error Rate (CER), and Sentence Error Rate (SER) of 4.02%, 0.8%, and 15.14% respectively, on random street addresses. In addition to testing street addresses, we performed accent-based and manual decoding testing on the developed ASR system. Results indicate the need to develop and deploy custom ASR systems for better accent adaptation and application-specific coverage. The developed ASR system is integrated into the TPL Maps a mobile application. It is Pakistan’s first Large Vocabulary Continuous Speech Recognition (LVCSR) real-time system to provide Urdu-based voice-activated navigation services.


  • Open Access

    Demand Forecasting in Supply Chain Management for Rossmann Stores Using Weather Enhanced Deep Learning Model

    by Nameer Ul Haq Qureshi, Salman Javed, Kamran Javed, Syed Meesam Raza Naqvi, Ali Raza, Zubair Saeed

    Demand forecasting is one of the essential aspects of supply chain management, as it is linked with the financial performance of the organization. In the retail industry, it is essential to have more accurate forecasts to make suitable decisions. Therefore, the selection of the right forecasting method is considered vital and ideal to meet customer needs. More precisely, this research paper focuses on developing forecasting model for 1115 Rossmann stores located in Europe. Although, previously researchers have been working on developing models to forecast sales demand and to improve accuracy. However, it has been observed that few of the necessary conditions or situations were not being catered for in sales demand forecasting. Such as most researchers used univariate data of total sales for forecasting demand. The internal and external factors such as weather, promotional activity, location of the store, and holidays also play one of the primary roles when it comes to sales demand to forecast. Therefore, it is not specifically a univariate problem but a multivariate problem which have been analyzed in this research. In this research, multivariate dataset including weather variables, other important features have been used in predicting sales demand in supply chain management which helped to achieve better and reliable results. An enhanced deep learning model for sales Demand Forecasting using Weather Data (SDFW) is proposed using Gated Recurrent Unit (GRU) with Grid search. The proposed approach GRU with Grid search showed better performances as compared to previously suggested Long Short Term Memory (LSTM) model. Moreover, Gated Recurrent Unit (GRU) with Grid Search showed significant improvement in sales demand forecasting accuracy when considering weather-related data subsets. These findings will help the Rossmann retail industry in predicting the upcoming sales demand in a more efficient way, which will also optimize their inventory records.


  • Open Access

    Real Time Worker Stress Prediction in a Smart Factory Assembly Line

    by Hassan Hijry, Syed Meesam Raza Naqvi, Kamran Javed, Omar H. Albalawi, Richard Olawoyin, Christophe Varnier, Noureddine Zerhouni

    This research contributes to a innovative approach to address the increasing issues of workplace mental health and stress, particularly in high-pressure environments like assembly lines which also affects workers performance and companies productivity. Recognizing the harmful effects of stress on worker productivity, this study introduce stress-monitoring model using advanced machine learning techniques. Proposed model integrates Internet of Things (IoT) technology and machine learning techniques, utilizing a wearable watch to gather open-source physiological data indicative of workers stress in assembly lines. Key physiological markers, such as heart rate, respiration rate, and skin conductance, are analyzed. Based on these physiological indicators, the primary objective is to develop and validate framework that can accurately predict worker stress levels using IoT and machine learning models. The empirical results of the proposed approach demonstrate that the most effective model achieves an impressive accuracy score, with the XGBoost model providing 99% accuracy and Matthew’s correlation coefficient (MCC) of 0.99, surpassing the performance of Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), k-Nearest Neighbors (kNN), and Support Vector Machine (SVM). The practical implications of these findings suggest a significant potential for implementing such technology in high-stress work settings, offering a proactive tool for stress management and contributing to enhanced worker well-being and productivity.