Using Google Speech API Transfer a Voice File to Text

Setting

• Create a new project on Google Cloud, like speech_project
• In the left side bar, select APIs & services -> dashboard
• In the APIs & services channel dashboard 
• Click the button ENABLE APIS and SERVICES
  
• Click Google Cloud Machine Learning - Speech API, in the new page Enable it.
• In the left side bar, select Credentials. Create a new API credential. Google suggests Service account key which could be used in different APIs. 
• 
• 
  
• Storage and Compute Engine setup below are not nessessary.
• Go to project's Storage Management
• Create a bucket, for example, named speech_bucket
• Upload voice files to this bucket
• Set up all files share publicly for simple. Otherwise, you need to deal with complicated certification issues.
• Go to COMPUTE ENGINE
• Create a new instance, like speech_vm
• Install Google Cloud SDK in a local computer.
• Connect to the instance speech_vm by this command gcloud compute ssh speech_vm

API usage

• Create a request config file, like sync-request.json
• The content in this config file looks like
  

  {
  "config": {
   "encoding":"LINEAR16",
   "sample_rate": 16000,
   "languageCode":"en-US"
  },
  "audio": {
   "uri":"gs://speech2text/test.wav"
  }
  }

• if the length of voice is longer than 1 minute, LINEAR16 is the only choice in encoding
• For a small voice file, the data could store in the HTTP request
• For a big voice file, you need to store it in cloud storage.
• Follow this document https://cloud.google.com/speech/reference/rest/Shared.Types/RecognitionConfig to improve recognization result.

• 调用
• Call speech API 
• curl -s -k -H "Content-Type: application/json" https://speech.googleapis.com/v1beta1/speech:asyncrecognize?key={YOUR_KEY} -d @sync-request.json
• I used asynrecognize method because the voice length is longer than 1 minute.
• You'll get a json file as response. The content in it is

  {
  "name": "5029905118673338253"
   }

• Use the name value to GET the status of speech recognization process

  https://speech.googleapis.com/v1beta1/operations/{NAME}?key={YOUR_KEY}

• You will get another JSON file from the URL above. The content in this JSON file would change until “done”=true

  {
  "name": "5029905118673338253",
  "metadata": {
  "@type": "type.googleapis.com/google.cloud.speech.v1beta1.AsyncRecognizeMetadata",
  "progressPercent": 100,
  "startTime": "2016-08-11T01:39:31.732812Z",
  "lastUpdateTime": "2016-08-11T01:57:25.101008Z"
  },
  "done": true,
  "response": {
  "@type": "type.googleapis.com/google.cloud.speech.v1beta1.AsyncRecognizeResponse",
  "results": [
   {
     "alternatives": [
       {
         "transcript": "I can share with you if that's okay",
         "confidence": 0.68820089
       }
     ]
   },

• The value of the key results is the final result