Deep Learning on Modulation Classification using Google’s Teachable Machine

Wireless communication is essential to people's daily lives and jobs. In the development of cognitive radios, automatic modulation classification (AMC) is a crucial and necessary topic. Finding the modulation schemes of unknown received signals is the goal of AMC, which is a stage between signal receiving and demodulation. Teachable Machine (TM) is a model development tool that uses Deep Learning (DL) and doesn't require any programming language. Google introduced Teachable Machine, a web-based artificial intelligence tool service for educators, learners, data scientists, and researchers who lack technical coding knowledge. An image dataset of 24,460 images using eight classes of digital modulation techniques—2PSK, QPSK, 8PSK, 16PSK, 8QAM, 16QAM, 32QAM, and 64QAM—is used to test the model created by TM. To enable us to train a model and subsequently assess its performance on unseen data, TM automatically divides the entire number of images into 85% for training and 15% for testing. The performance of the suggested model is examined experimentally using receiver operating characteristic (ROC) curves and classification reports of modulation approaches. According to the success rate statistics, 3,417 out of 3,669 images passed the generated model's accuracy test with 94.19 percent.