Final Year Project UNIKL BMI

During week 3 of the project, my focus shifted towards the design of the Artificial Neural Network (ANN) model for the Malaysian Sign Language Recognition System. To accomplish this, I made the decision to utilize an ANN as the machine learning model, which is illustrated in the provided image. The ANN architecture encompasses several key components, including an input layer, two hidden layers, and an output layer. The input layer is composed of 42 neurons, which receive the relative x and y coordinates of the 21 keypoints extracted from the MediaPipe hand landmark model. These coordinates serve as the fundamental input data for the ANN model. Moving on to the hidden layers, the first hidden layer consists of 20 neurons, while the second hidden layer is comprised of 10 neurons. Both of these hidden layers employ the Rectified Linear Unit (ReLU) activation function, a non-linear function that aids in capturing complex patterns and relationships within the data. The output layer of the A

In the second week of my project development, I dedicated my efforts to the crucial tasks of processing and cleaning the data that was collected in the preceding week. The data processing stage was essential to ensure that the gathered data could be effectively utilized for training the model later on. This involved performing various operations, such as formatting, organizing, and transforming the data into a suitable structure that aligns with the requirements of the model. Simultaneously, the data cleaning stage played a vital role in preparing the dataset for analysis. This involved meticulously examining the data for any inconsistencies, inaccuracies, or abnormalities that could potentially impact the performance and reliability of the model. In the process of data cleaning, any identified issues were addressed and rectified to ensure the data's integrity and quality. Subsequently, after the processing and cleaning stages were completed, the data was split into two distinct su

 In the first week of FYP 2, I had decided to develop the code for hand detection that will be use during the data collection process. The code involve libraries such as OpenCV and Mediapipe's hand detection solution that helps in detecting the hand and mapping data points on the hand. The output generated are as the image below.   During the coding process, I stumbled upon a problem which is to create an output window that only capture the hand part. This process is important because it helps in improving the sign recognition task by reducing the size of the image and cropping the unwanted background that could affect the quality of the generated sign recognition algorithm later. i continue the make research on how to solve this problem in several platform such as youtube, stackoverflow, and chatGPT.