Accelerating the Value of In-Car Speech Systems

We all dream about being able to tell our cars to find the quickest way to grandmother’s house while playing the early hits of the Beatles. 

 On the way, you tell the car you would like to find the closest doughnut shop with the highest-rated soy latte. Before you know it, there you are, going over the mountains and through the woods, sipping a latte, singing along with “Can’t Buy Me Love,” without ever taking your hands off of the wheel.

Much of this dream is now something of a reality: Top automobile manufacturers, such as Mercedes-Benz, Ford, Hyundai-Kia, Toyota, Lexus, and BMW, all offer voice control options in their cars. In early June, Ford released an updated version of its SYNC system, which uses technologies from Microsoft/Tellme and Nuance Communications, to enable drivers to use voice commands to get directions, find movie times, check stocks, and even hear a horoscope reading.

It would seem that when it comes to the car, voice controls are quickly becoming ubiquitous because many consumers see the possibility for both convenience and safety. However, there have been concerns that continue to arise around these technologies. Strategy Analytics’ recent Automotive Consumer Insights (ACI) report states the problems with speech technology are very different from problems users face when dealing with other inputs, such as touchscreens or buttons.

“While speech as a user interface has come a long way in terms of accuracy for simple tasks, the promise of the voice-controlled connected car with more complex features may negate these advances,” commented Chris Schreiner, senior analyst at Strategy Analytics, a firm that often focuses on automotive electronics and entertainment. 

Unfortunately, while speech technologies are in demand—79 percent of drivers view them as desirable—the report shows that user satisfaction was as low as 33 percent when it came to the Nokia HF-33W, and didn’t reach higher than 70 percent on the Parrot MKi9200 that is often used by Lexus owners. Sony Ericsson HCB-108 and Motorola T505 came in somewhat lower, at slightly less than 60 percent.

One of the central problems of voice technology interfaces, Schreiner points out, is natural language speech. “Natural language speech is going to be lower [in accuracy] than command and control, but the other issue is that consumers don’t know when to speak and talk over the prompts.”

Furthermore, Schreiner says marketing is a concern among companies trying to sell voice technologies for the automotive industry. As the study reports, ease of use is crucial to maintaining safety. “I’ve done research that shows that there is a safety benefit until there are system errors, and the more cognitive attempts are used in correcting the error, the more problems [during] driving can occur,” Schreiner says. 

To help improve these issues, ACI makes the following recommendations:

• keep manual interaction to a minimum;
• provide short but informative prompts;
• provide useful help;
• provide assistance for novice users and adapt for experts;
• design a sensible interruption strategy;
• limit voice functionality to moderate and complex multimedia tasks; and
• provide simple error recovery.

Overall, ACI recommends that vendors employ speech via a hybrid strategy that combines embedded and server-side recognition engines.