<![CDATA[Emotion is a feeling that causes behavioral and interaction changes towards one's surroundings. One way to detect human emotions is through speech. According to a survey, it was found that 1.6 million Indonesian teenagers experience anxiety disorder (Erskine et al., 2021). Anxiety disorder is a type of anxiety characterized by feelings of fear or vigilance that are not clearly defined (Saleh, 2019). Therefore, this study aims to create a tool to detect a person's emotional condition through voice processing, specifically fear emotions based on three intensity levels: low, medium, and high, with the purpose of serving as a consideration for psychologists/psychiatrists in the initial screening and diagnosis stages.. This research uses Gammatone-Frequency Cepstral Coefficients (GFCC) extraction method which has an effective gamma filter for speech with high noise. In addition, this research also tests the capability of Random Forest Classifier classification in recognizing the intensity of fear emotion from speech signal. This research is important because it can provide information on the effectiveness of GFCC towards noise and the accuracy of GFCC extraction in the intensity detection system of fear emotion. This system is developed using a Raspberry Pi 4B as the processor and connected to an Android application for displaying the classification results. The connection between the Raspberry Pi 4B and the application is established via Bluetooth using the RFCOMM communication protocol. This research concludes that Signal-to-Noise Reduction on speech processed with GFCC extraction is more effective compared to Mel-Frequency Cepstral Coefficients (MFCC). The accuracy of the implementation of the emotion level recognition system using GFCC extraction and Random Forest Classifier classification is 73.33%.]]>
