Is there a Google Photos API?

Why does "Google web speech API" not recognize automated attendant unlike real voice or recorded voice?

• Google web speech API URL:http://www.google.com/intl/en/chrome/demos/speech.html I'd like to transfer automatically automated attendant to text by using Web speech API. I've already tried it, but it didn't seem to work well. As I use Google web speech API, it recognizes real voice and recorded voice. When I spoke on microphone by using Google web speech API, it recognized my voice and transferred speech to text well. Also I played one of ted video on my phone and brought the phone close to microphone on mac book. It was also well transferred from recorded voice to text, even though speaking speed was very fast. But when I brought the phone with speakerphone mode while it is emanating "automated attendant", Google web speech API didn't recognize most of word in it. Does anybody know why web speech API doesn't recognize automated attendant?

• Answer:

  You probably want to provide a recordings of the sounds you are investigating to get a more definite answer. One very possible reason is a frequency band mismatch. Ted talks and your speech are wide-band audio with full range of frequencies up to 8khz. The telephone attendants are optimized to telephone lines and they have only frequencies up to 4khz. Google speech API might not handle that cut well because ASR is not very robust to frequency cut these days. To recognize telephone speech you might want to use specialized models like the ones provided by http://cmusphinx.sourceforge.net