fbpx

Offline Speech Recognition System for Drivers

20 Thousand Leagues took up building a speech recognition as part of an Advanced Driver-Assistance System. The solution is implemented in cars that do not have ADAS of their own to keep otherwise fine car relevant in B2B and B2C cases for a few more years.

Challenge

A car without speech recognition means a car without Advanced Driver-Assistance System or a car with poor ADAS. As a result, we do not have (enough) in-built processing power to simply install a speech recognition system onto the existing hardware. An external device is required to implement our solution.

A lot of speech recognition systems partially or fully rely on cloud to recognize the input and/or process that. This is a suboptimal solution because drivers do not always have access to the internet, especially in rural areas or developing countries. Besides, server infrastructure from AWS and others has to catch up with the demand every day. Still, the transfer gets more and more cluttered on the ISP’s side.

We need a small yet powerful external device to install speech recognition on older cars. And make it without online components for longevity’s sake.

Single-board computer is both powerful and tiny. Our proprietary speech recognition system doesn’t need internet.

Drivers don’t have to take their eyes off the road. Older cars stay relevant for a few more years.

Solution

External hardware for speech recognition should not physically interfere with the car manufacturer’s dashboard. The NVIDIA Jetson single-board computer fits perfectly, even for smaller cars such as Smart Fortwo.

Before this project, 20kL had put two years into building an offline speech recognition system. It was optimized for the single-board computer. The computer comprises both the dictionaries and the algorithm for stacking input against all words in the dictionary. The system was executed in Python.

Building software is a team sport. Without proper coordination, communication and verification; you’ll end up with a sub-par product and a whole bunch of people dissatisfied. This is also why we at 20kL are frank, upfront, transparent and direct with our customers.
R9707ci8TTSLAfl0Ct8mug

Simon Kaastrup-Olsen
Chief Executive Officer

Results

Just like our object detection system for cars, the solution was ordered by an IT company rather than a car manufacturer and has been further redistributed to other businesses. This includes a large car-sharing company, as entry-level cars often do not have the processing power for their own speech recognition systems.

Technology: Speech Recognition

Tools and framework: Python

We want to build the best software we can, we do that by working with our customers by asking lots of questions, visualising every possible bit, transparent coordination and no delayed decisions
R9707ci8TTSLAfl0Ct8mug

Simon Kaastrup-Olsen
Chief Executive Officer

Next story

Repair Automation

Rope Robotics