The vulnerability of the speaker identity verification system to attacks using voice cloning was examined. The research project assumed creating a model for verifying the speaker’s identity based on voice biometrics and then testing its resistance to potential attacks using voice cloning. The Deep Speaker Neural Speaker Embedding System was trained, and the Real-Time Voice Cloning system was employed based on the SV2TTS, Tacotron, WaveRNN, and GE2E neural networks. The results of attacks using voice cloning were analyzed and discussed in the context of a subjective assessment of cloned voice fidelity. Subjective test results and attempts to authenticate speakers proved that the tested biometric identity verification system might resist voice cloning attacks even if humans cannot distinguish cloned samples from original ones.

An approach to a new kind of recommendation system design that suggests safe speed on the road is presented. Real data obtained on roads were used for the simulations. As part of a project related to autonomous road sign development, a number of measurements were carried out on both local roads and expressways. A speed recommendation model was created based on gathered traffic data employing the traffic simulator. Depending on the traffic volume and atmospheric conditions prevailing on the road, as well as the surface conditions, the proposed system recommends the safe speed for passing vehicles by influencing the distance from the preceding vehicle to prevent collisions. The observed effect of the system application was an increase in the minimal distance between vehicles in most simulations.

The article presents preliminary experiments investigating the impact of accent on the performance of the Whisper automatic speech recognition (ASR) system, specifically for the Polish language and medical data. The literature review revealed a scarcity of studies on the influence of accents on speech recognition systems in Polish, especially concerning medical terminology. The experiments involved voice cloning of selected individuals and adding prosodic contours with Russian and German accents, followed by transcription of these samples using all available models from the Whisper family and comparison with the original transcription. The results of these initial experiments suggest that the Whisper model struggles with foreign accents in the context of Polish language and medical terminology. This highlights the need for further research aimed at improving ASR systems for transcription of medical personnel.

Identifying different vehicle types can help manage traffic more efficiently, reduce congestion, and improve public safety. This study aims to create a classification model that can recognize vehicle types based on the sound of passing vehicles. To achieve this, a database of raw audio files containing 1763 samples from two sources was assembled. The time-domain signals were converted to a time-frequency representation using the short-time Fourier transform to generate Mel Spectrograms. Mel-frequency Cepstral Coefficients (MFCCs) were also generated using the discrete cosine transform. In our experiments we compared these approaches. Since the data was imbalanced we applied online augmentation. Based on the literature review, we chose a Convolutional Neural Network (CNN) classifier because it is particularly well suited for analyzing large datasets due to its automatic feature extraction, parameter sharing and sparsity. The results showed that Mel Spectrograms were more effective for audio data preprocessing in this particular use case, achieving the highest accuracy of 0.875 and the highest f1-score of 0.877 compared to MFCCs.