The recognition of voice commands in high-noise driving environments presents unique challenges that require implementing advanced strategies to ensure reliable and accurate performance. This paper introduces a method to improve the accuracy of voice command recognition in those environments. This study addresses the need for advanced strategies in the development of robust speech recognition systems for underresearched languages, such as Brazilian Portuguese, by optimizing speech-to-text results and introducing a new tool, the Audio Dataset Creator. It simplifies the creation of audio datasets and automatically validates all recorded audio through autogenerated transcripts. The study also emphasizes the potential impact of voice command recognition for driver assistance, particularly in enhancing accessibility and safety in vehicles adapted for drivers with reduced mobility. A novel Brazilian Portuguese audio dataset for vehicle dashboard control was created to develop the voice assistant system. This method demonstrated superior performance when compared to other speech recognition tools, correctly recognizing up to 99% of the voice commands, with only a small set of data for training. Furthermore, our transcriptbased command estimation approach shows an average 30% improvement in recognition accuracy. These results demonstrate the efficacy of the proposed method and tools in advancing voice command recognition technology for onboard automotive applications, which require lightweight and efficient solutions.