Аннотация
The study aims to develop a neural network for voice modification in real time. The main attention is paid to the issues of audio signal capture and processing, feature extraction using methods such as MFCC (Mel-frequency cepstrum, Low-frequency cepstrum coefficients), as well as the development and training of a neural network model based on LSTM (Long Short-Term Memory) to effectively change voice characteristics (timbre, pitch, speed). The methodology includes the use of machine learning techniques to create a flexible and intuitive system with minimal delay in signal processing. Optimizing the system for real-time operation and ensuring high-quality modified sound are key research objectives. To achieve these goals, it is planned to analyze existing approaches to audio signal processing and voice modification. Various neural network architectures, including convolutional and recurrent models, will be considered to select the most appropriate solution. Special attention will be paid to data collection and preparation, including recording and annotating audio files with various voice characteristics. In conclusion, it is planned to test the system on real data and evaluate its performance in terms of sound quality and processing speed