Speech Data Collection for In-Cabin Recognition

The Challenge:

A leading automotive software provider approached us to collect speech data, addressing the challenge of improving their in-cabin speech recognition capabilities. Building a system that accurately understands and processes voice commands necessitated a diverse and extensive dataset. However, variations in speech patterns, including dialects, accents, and tones, posed a significant obstacle. Moreover, the complex background noise within the vehicle affected the recognition system's accuracy.

The Solution:

Our comprehensive services encompass natural speech data collection, covering various scenarios and linguistic variations. Guided by our language experts, we efficiently gather speech data in different languages and locations, tailored to specific scenarios. Utilizing professional equipment, we simulate diverse environmental conditions, capturing authentic interactions without scripted dialogue. We involve native speakers from multiple countries, ensuring accurate representation of drivers' emotions and other essential elements within the automotive industry's language scope.

The Result:

Through our data collection and annotation services, the company successfully expanded its system's support for over 30 languages, significantly enhancing their in-cabin speech recognition capabilities.