Speech Recognition Datasets

This dataset contains 531 hours of in-car ambient noise recordings captured using microphones and mobile phones across various vehicle models, road types, speeds, and cabin conditions such as windows open or closed. The noise was recorded at six distinct points inside each vehicle to reflect spatial diversity and better support vehicle sound modeling. The dataset captures real-world driving environments, engine hums, road interactions, wind noise, and cabin reverberation. It is ideal for use cases such as noise suppression, automatic speech recognition (ASR) in cars, in-vehicle audio enhancement, and sound source separation. Validated by leading AI companies, the dataset complies fully with global data privacy regulations including GDPR, CCPA, and PIPL, making it suitable for both research and commercial applications.

This dataset contains 19 hours of real-world ambient noise recordings captured in bus environments using the Tascam DR-07x voice recorder. The audio was collected from both in-bus scenes and bus platform areas under various real-life conditions. The dataset reflects authentic public transportation soundscapes including engine hum, passenger chatter, door movement, platform announcements, and other background noises. It is suitable for a wide range of applications such as acoustic scene classification, noise suppression, automatic speech recognition (ASR) in noisy environments, and audio enhancement models. The dataset has been tested and validated by leading AI companies and adheres strictly to global data protection standards including GDPR, CCPA, and PIPL, ensuring compliance and safe use in commercial and research applications.

This dataset contains 1,297 hours of environmental noise recordings collected using voice recorders across diverse real-world scenarios, including subways, supermarkets, restaurants, roads, and more. All recordings are annotated with timestamps and relevant metadata, making the data ideal for training AI models in noise reduction, environmental sound classification, audio preprocessing, and speech enhancement tasks. The dataset has been validated by leading AI companies and complies with global data protection regulations including GDPR, CCPA, and PIPL.

This dataset contains 101 hours of environmental noise recordings captured using high-quality voice recorders across a wide range of real-world locations, including subways, supermarkets, dining halls, city streets, airports, cinemas, expressways, and high-speed trains. It offers diverse ambient sound conditions ideal for tasks such as noise reduction, environmental sound classification, and audio preprocessing in speech or sound recognition systems. The data has been validated by major AI companies and complies with data protection regulations including GDPR, CCPA, and PIPL.

This 10-hour Far-Field In-Home Noise Dataset was collected using multiple types of microphone arrays installed in real family home environments. Each mic array setup offers varied spatial capture perspectives, making the dataset ideal for AI tasks such as far-field automatic speech recognition (ASR), voice enhancement, smart speaker training, and multi-microphone signal processing. All data has undergone rigorous quality validation and complies with global privacy regulations including GDPR, CCPA, and PIPL.

This dataset contains 20 hours of radio frequency noise audio recorded via high-quality microphones in 66 different rooms, with 2–4 recording points per room and multiple recording angles per point. The setup simulates real-world RF interference and ambient indoor noise scenarios, supporting tasks like sound source localization, acoustic modeling, and noise classification. The dataset has been validated by leading AI firms and complies with all major privacy regulations, including GDPR, CCPA, and PIPL.

Mandarin Chinese(China) Noisy Monologue Smartphone speech dataset_ Guiding, collected from monologue based on given prompts, covering generic domain, such as in-car, smart home, voice assistant, recorded in noisy condition. Transcribed with text content and other attributes. Our dataset was collected from extensive and diversify speakers(205 people), geographicly speaking, enhancing model performance in real and complex tasks. Quality tested by various AI companies. We strictly adhere to data protection regulations and privacy standards, ensuring the maintenance of user privacy and legal rights throughout the data collection, storage, and usage processes, our datasets are all GDPR, CCPA, PIPL complied.