 |
Siemens VDO Presents Speech Recognition for the Automobile
9th June, 2006
Humans engage in dialogue very early on and the spoken language is considered the most natural means of communication. With speech recognition, Siemens VDO is making a contribution towards using this form of communication as the interface between man and machine (Human Machine Interface, HMI) in the automobile, as well, enabling voice control to help accelerate the operation of various functions. The automobile supplier is providing an integrated HMI concept and intends operation of complex functions through a sensible combination of voice commands, haptic and visual elements and handwriting recognition (EasyCo). Currently, Siemens VDO speech recognition can address 70,000 city and street names. By 2012, the system is planned to be able to process up to 300,000 words.
After learning to read, the automobile is learning to speak. In the future, drivers will be able to choose their preferred way to communicate with their cars. The challenge for the automobile supplier is to develop speech recognition that is as close as possible to natural language in expression and syntax. Dialogue between humans and machines that is not restricted by fixed commands will become a reality in the future. What started with the EasyCo system, the recognition of handwriting using a touchpad, Siemens VDO is now continuing with speech recognition.
The car's first “lesson” will be to learn words, as the average, actively used vocabulary, for example in German, includes about 10,000 words. In comparison, the German Duden dictionary contains about 130,000 entries. Above all, a larger vocabulary is a prerequisite when entering navigation information. Siemens VDO speech recognition currently can directly address more than 70,000 German city and street names. Twelve languages, including Chinese and Japanese, can be selected with more on the way.
The basic principle of the speech recognition's functionality begins with the speech signal being received by a microphone, filtered of ambient background noise and the unique software searches for key terms. The terms are associated to the audio signal with a semantic command from which the action, for instance selecting a name from a telephone book, can be derived.
A significant advantage of speech recognition is the ability to directly call up sub-functions through Global Commands without having to go through various menu levels. The number and variety of target functions are arbitrary. For drivers, this means increased safety, along with simplified operation, as they can better concentrate on traffic.
As an interior, cockpit and information solutions supplier, Siemens VDO is developing an integrated HMI concept intended for the operation of complex functions through a sensible combination of haptic and visual elements, voice commands and EasyCo handwriting recognition. With this comprehensive approach, Siemens VDO is supporting automobile manufacturer requirements place on ergonomics and operating safety and allows the customer to decide on as many operating options as possible, depending on the traffic situation and the driving preferences.
In addition, particular commands also will be accessed by voice. For example, a passenger can find out the evening’s cinema program while traveling to the theater. Dynamic dialogue is the key element of future speech recognition between the user and the system turning it into fluent communication more closely resembling the natural speech. In the long term, Siemens VDO is working on improved future recognition of each individual term. An image recognition sensor for reading lips and specially arranged microphones focused in the speaker's immediate proximity could be used to improve recognition. As a result, the commands can be clearly allocated to the speaker – even if that person is not the driver. A larger vocabulary is being worked on in parallel to this. For the automobile, that means learning approximately an additional 230,000 words by the beginning of the next decade.
Url:
http://www.autospectator.com/modules/news/article.php?storyid=4710
|
|
