Add audio_raw_to_speech, basics of gCloud Speech API

bin/audio_raw_to_speech.rb

@@ -0,0 +1,76 @@
+#!/usr/bin/env ruby
+###################
+
+#
+# Load the library path
+#
+base = __FILE__
+while File.symlink?(base)
+  base = File.expand_path(File.readlink(base), File.dirname(base))
+end
+$:.unshift(File.join(File.expand_path(File.dirname(base)), '..', 'lib'))
+
+require 'warvox'
+
+require 'uri'
+require 'net/http'
+require 'json'
+
+def usage
+  $stderr.puts "Usage: #{$0} <input.raw> <output.json>"
+  exit
+end
+
+#
+# Script
+#
+
+inp = ARGV.shift
+out = ARGV.shift
+
+if (inp and inp == "-h") or not inp
+  usage()
+end
+
+raw = WarVOX::Audio::Raw.from_file(inp)
+res = nil
+flac = raw.to_flac
+akey = WarVOX::Config.gcloud_key
+
+if ! akey
+  $stderr.puts "Error: A gcloud API key needs to be configured"
+  exit(1)
+end
+
+uri = URI('https://speech.googleapis.com/v1/speech:recognize?key=' + akey)
+req = Net::HTTP::Post.new(uri, initheader = {'Content-Type' =>'application/json'})
+
+loop do
+  req.body =
+  {
+    "initialRequest" => {
+      "encoding"     => "FLAC",
+      "sampleRate"   => 16000,
+    },
+    "audioRequest" => {
+      "content" => [flac].pack("m*").gsub(/\s+/, '')
+    }
+  }.to_json
+
+
+  http = Net::HTTP.new(uri.hostname, uri.port)
+  http.use_ssl = true
+  res = http.request(req)
+
+  break if res.code.to_s == "200"
+  $stderr.puts "Retrying due to #{res.code} #{res.message}..."
+  sleep(1)
+end
+
+if out
+  ::File.open(out, "wb") do |fd|
+    fd.write(res.body)
+  end
+else
+  $stdout.write(res.body)
+end

bin/verify_install.rb

@@ -56,6 +56,14 @@
 puts "[*] The LAME binary appears to be available"
 
 
+if(not WarVOX::Config.tool_path('sox'))
+  puts "[*] ERROR: The 'sox binary could not be installed"
+  puts "[*] $ sudo apt-get install sox"
+  exit
+end
+puts "[*] The SOX binary appears to be available"
+
+
 puts " "
 puts "[*] Congratulations! You are almost ready to run WarVOX"
 puts " "

config/warvox.conf

@@ -8,8 +8,10 @@
 tools:
   gnuplot: gnuplot
   lame: lame
+  sox: sox
   iaxrecord: "%BASE%/bin/iaxrecord.rb"
 
+
 #
 # Maximum processing jobs, normally this is set to your processor core count,
 # but you can limit it further here. Keep in mind that each analysis job also
@@ -32,3 +34,9 @@ classifiers: "%BASE%/config/classifiers"
 # Configure the signature directory
 #
 signatures: "%BASE%/config/signatures"
+
+#
+# Configure cloud integrations
+#
+apis:
+  gcloud: "%BASE%/config/gcloud.key"

lib/warvox/audio/raw.rb

@@ -1,3 +1,5 @@
+require 'open3'
+
 module WarVOX
 module Audio
 class Raw
@@ -72,6 +74,16 @@ def to_wav
       raw
   end
 
+  def to_flac
+    sox = WarVOX::Config.tool_path('sox')
+    if ! sox
+      raise RuntimeError, "The sox binary could not be find, make sure it is installed"
+    end
+
+    o, s = Open3.capture2("#{sox} -t raw -b 16 -e signed-integer -r 8000 - -t flac -r 16000 -", :binmode => true, :stdin_data => self.to_raw)
+    o
+  end
+
   def to_flow(opts={})
 
     lo_lim = (opts[:lo_lim] || 100).to_i

lib/warvox/config.rb

@@ -49,6 +49,14 @@ def self.blacklist_load
 
   end
 
+  def self.gcloud_key
+    info = YAML.load_file(WarVOX::Conf)
+    return nil if not info
+    return nil if not info['apis']
+    return nil if not info['apis']['gcloud']
+    ::File.read(File.expand_path(info['apis']['gcloud'].gsub('%BASE%', WarVOX::Base)))
+  end
+
   def self.signatures_path
     info = YAML.load_file(WarVOX::Conf)
     return nil if not info