By Mark Moeller
One of the cornerstone features of the Auto PC is its speech driven user interface. In an effort to reduce distraction to the operator of the vehicle and to keep the operator's hands on the wheel and eyes on the road, speech recognition seems a natural answer. Speech is the only open ended method of "output" that the driver of the car has left.
The ideal speech user interface would allow you to talk to your Auto PC just as you would talk to any passenger in your car. Theoretically, it would understand you as well (or better) than some passengers would. There are many reasons that this technology simply doesn't exist today. Yes, you will see some pretty amazing demos of speech recognition systems with natural language parsing on the back end, but these are primarily research projects running on very powerful and expensive CPUs.
One of the biggest stumbling blocks to a successful speech recognition user interface is simply identifying who or what the computer should be listening to. The microphone that comes with the Auto PC is a simple single element, wide pickup device. All incoming audio gets treated equally and it's left for the speech recognition engine to sort out speech from the incoming cacophony of sound. In a noisy environment, this is difficult at best, but usually impossible.
Compare this with the auditory system that people have. Our two ears feed processed signals to the most powerful computing device known to mankind, the brain. The brain then does some non-trivial signal processing, measuring the arrival time of a given sound to each of the ears. Correlating the incoming signals and presenting the hearer with an array of sounds, we can then individually focus on each and know where the sound originates in three dimensions. This is obviously oversimplified, but illustrates the advantage our auditory system has over a single microphone. If you enjoy experiments, try identifying where three individual speakers are located at with respect to each other when listening to a recording of their speech as they are all talking.
Recently, a company by the name of Andrea Electronics has developed a microphone for the Auto PC that effectively gives the Auto PC four "ears" and the ability to focus on a single speaker.
Andrea AutoArray DA-300
The product is called the Andrea AutoArray DA-300 and it's made specifically for the Auto PC. This microphone, shown in Figure A, replaces the stock "mouse style" microphone that ships with Clarion's AutoPC 310C. It is a phased array microphone that is specifically designed for enhancing speech recognition in an automobile. This microphone operates as four separate microphone elements, which run their signals into a digital signal processor. The signal processor then adaptively focuses the microphone on the driver's voice and rejects other noise.
