With the rapid advancement of artificial intelligence technologies in human life, particularly within the automotive industry, the popularity of smart vehicles has increased. Designing an intelligent car cabin capable of seamless interactions with both the driver and the vehicle emerges as a crucial solution to address human-vehicle interaction challenges. This project aims to implement a digital voice assistant that recognizes vehicle commands. It utilizes three main techniques: speech-to-text conversion, text classification, and calibrating the classifier in order to detect out-of-distribution (OOD) sentences. Using Vosk, an LLM model, voices in the vehicle environment are converted into text format. Then, after pre-processing the text, an SVM classifies them. Using Platt scaling the SVM classifier outputs become calibrated, which makes them probabilistic. The experimental result shows that the proposed model achieves command recognition with 96.43% accuracy, 96.83% precision, 96.43% recall, 96.39% F1-score, and 0.1437 cross-entropy loss. Furthermore, the optimal threshold for OOD sentences is 0.4.
Colloquial Bengali has adopted many English words due to colonial influence. In conversational Bengali, it is quite common to speak in a mixture of English and Bengali, a phenomenon termed Code-switching (CS). To build a Voice Command Classifier in this era, when the usage of CS is ever-increasing, it is often necessary to map a single base command to its many different variants - spoken in multiple mixtures of languages. The works done with Bengali Speech have been primarily focused on single word classification and mostly incompetent in understanding the complex semantic relationships displayed in sentences. This paper proposes ‘CSVC-Net’, a CNN-LSTM based architecture for classifying spoken commands that exhibit code-switching between Bengali and English. To effectively reflect the scenario, it also presents a newly curated dataset named ‘Banglish’ containing 3,840 audio files of spoken computer commands belonging to 11 classes, considering 64 variations in total. The proposed pipeline passes the input audio signal through a series of appropriate transformation and augmentation steps enabling the model to achieve an accuracy of 92.08% on the curated dataset. Furthermore, the robustness of the proposed model has been justified by comparing with different architectures and tested under different noise levels with promising accuracy, which shows the applicability of the model in real-life scenarios.
Voice-enabled commercial products are ubiquitous, typically enabled by lightweight on-device keyword spotting (KWS) and full automatic speech recognition (ASR) in the cloud. ASR systems require significant computational resources in training and for inference, not to mention copious amounts of annotated speech data. KWS systems, on the other hand, are less resource-intensive but have limited capabilities. On the Comcast Xfinity X1 entertainment platform, we explore a middle ground between ASR and KWS: We introduce a novel, resource-efficient neural network for voice query recognition that is much more accurate than state-of-the-art CNNs for KWS, yet can be easily trained and deployed with limited resources. On an evaluation dataset representing the top 200 voice queries, we achieve a low false alarm rate of 1% and a query error rate of 6%. Our model performs inference 8.24x faster than the current ASR system.