Browsing by Author "Waseem"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Clutter and jammer mitigation for airborne radars
    (UMT.Lahore, 2018) Muhammad Hassan; Qazi Muhammad Raheel; Waseem; Shahzad Khalil
    Detection of signal in the presence of noise was topic of great interest in the domain of Radar signal processing since after 2nd world war. But in that scenario the platform (Ground) itself was stationary and it was not so challenging to detect the signal in the presence of clutter or noise because clutter filtering can be easily done on the basis of Doppler shift .But in the last 20 years radar world has changed dramatically and active multichannel phased array antennas are now mostly using in every fighter plane mostly known as AESA (Active Electronically Scanned – Array radars).These Airborne radars have the capability to target any stationary or moving object one the ground . In case of targeting any moving object on the ground although the one problem is that the signal is totally embedded in noise. The second major problem is that Doppler frequency of clutter is dependent on the angle between clutter and radial velocity of Aircraft. Space time adaptive processing is the technique which filters the signal in both the spatial and temporal domain. For spatial sampling of the backscattered echo there must be multichannel phased array antenna, so the array geometry comes on account of spatial dimensions. And the 32temporal dimensions come on account of pulse trains. Adaptive processing is derived from the word 'Adaptation’. This means on account of changing nature of clutter and jammer noise there is a need to estimate the "inverse" of clutter and jammer noise space time covariance -matrix. For modeling all of these STAP algorithms current program must have to link a complete mono-static radar. So first modeling a complete mono-static radar in which based on radar design specifications radar parameters will be designed, target, transmitter, receiver and propagation environment effects will incorporate and After that linking this radar for performing space time adaptive processing.
  • No Thumbnail Available
    Item
    Sentiment analysis from manually annotated car reviews dataset using different machine learning approaches
    (UMT Lahore, 2021) Waseem; Muhammad Zubair; Talha Tariq
    Sentimental analysis is one of the creative research analyses in the computer science field. In later times, the analysts centered on identifying and isolating different features from users’ surveys, comments, and responses. According to our knowledge, there is no aspect-based categorized dataset available on vehicle reviews. We collect the vehicle reviews dataset from different resources. Then we generated the aspect-based categorized dataset. Third, preprocess the generated dataset and train models on it, and then perform comparative analysis on the results. The objectives of our work are to generate the dataset on the basis of aspects and sub-aspects. After generating the dataset, we can perform different machine learning techniques on it. Breaking a review on aspect and sub-aspect categories like aspect (performance), sub aspect (average, mileage), aspect (driving) sub aspect (handling, braking), etc. The latest evolutions in machine, deep and neural learning models demonstrate that they can help aspect-based sentiment analysis (ABSA) to figure out the awareness in human words and expressions. However, in the latest research, many ideas launched for ABSA generally focused on detecting the features, without taking into account the connection of such words with regard to the other aspects and features within sentences. That is the reason, these approaches are not adequate enough to yield decisions that contain more than one feature and sentence with difficult and certain constructions. As well, categorizing these viewpoints based on their assumptions has not been discussed in the past approaches. Besides, to the greatest of our knowledge no instrument or tool has isolated viewpoint term features and proposed a reputation chart that speaks to the shortcoming, best feature, and positioning of aspects based on features that are talked about within the audits in a single demonstration. To deal with this problem, first, we perform classification on the generated dataset using LazyPredict to find the best performance of the different classifiers. We use the classifiers with the best performance like LGBM classifier, random forest, XGB classifier, NuSVC, and extra tree classifier as proposed in the dataset. After analysis, the outcome of the experimental result of our proposed model will be contrasted to find the best approaches.