Improve Accuracy Intel Real Sense Speech Recognition

How can we improve accuracy in speech recognition. Currently i am using Commands and Dictionary. Are there any more ways to improve this further. Can we create an audio dictionary as well such that a particular audio should be recognized as a particular sentence.

 PXCMSpeechRecognition sr;
   session.CreateImpl<PXCMSpeechRecognition>(out sr);
   // sr is a PXCMSpeechRecognition instance.

            String[] cmds = new String[9] { "Switch on",  
                  "Light Off",
                 "Up",
                 "Down",
                 "Raise",
                 "Two",
                 "Three",
                 "Four",
                 "Five"
              };

            // Build the grammar.
            sr.BuildGrammarFromStringList(1, cmds, null);

            // Set the active grammar.
            sr.SetGrammar(1);

---

You can set the Volume up Take a look

I think if you set your volume up the system can recognize better.

As the man said to you , you have to limit your dictionary, but be careful if there are words with a seem pronunciation , maybe it may confuse the sound. EXAMPLE :

   public void recognize(){

        PXCMAudioSource audioSource = session.CreateAudioSource();
        audioSource.SetVolume(1f);
        ....

      }

---

One way to improve is to limit your dictionary to expected commands, phrases. That way instead of checking thru 50k words, it can be much smaller.

链接地址: http://www.djcxy.com/p/34366.html

上一篇: 如何收集和准备用于语音识别的数据？

下一篇: 提高英特尔真实语音识别的准确度