Case study for In-vehicle Speech

The company who is a leading automotive software provider approached us to help them collect speech data in order to meet the challenge of in-cabin speech recognition capabilities.

The Challenge

For in-cabin speech recognition system, to correctly recognize and process voice commands must be trained with a large amount of speech datasets. But the results can vary greatly depending on how people speak. Drivers may have various verbal commands to adjust the car’s temperature, volume, navigation, phone, or equipment adjustments. Training these speech video systems to understand multiple dialects, accents, and tones poses a significant challenge. The intricate environmental background interference in the vehicle dramatically affects the accuracy of the recognition system.

The Solution

·We provide natural speech data collection services covering all scenarios and variations. 

·With the guidance of our language experts, we can professionally and efficiently perform speech collection functions for each specific scenario and quickly for various languages in various locations. 

·We use professional equipment to simulate different scenario environments without any script and spontaneously perform voice data collection. 

·Native speakers from multiple countries are collected for each scene to carry out simulation restoration, the actual restoration of the driver's emotional state of mind and other elements, and then manage a variety of languages used in the automotive industry.

The Result

With our data collection and annotation services, the company has successfully expanded its system with 30+ languages.