extract solamimi from mp3 file

solamimi.rb

require 'rubygems'
require 'scissor'
require 'pit'

# = ABSTRACT
#   extract solamimi text from mp3 file using scissor, ffmpeg and julius
# = USAGE
#   ruby solamimi.rb /path/to/target.mp3 start_time length
# = EXAMPLE
#   ruby solamimi.rb ~/Music/iTunes/iTunes\ Music/The\ Beatles/1962-1966\ \[Disc\ 1\]/1-02\ Please\ Please\ Me.mp3 ~/20100103/dictation-kit-v4.0/fast.jconf 1:02.22 2.3

class Solamimi
  def self.listen_to(mp3_path, options=nil)
    self.new(mp3_path, options).process
  end

  attr_reader :japanese, :start, :length

  def initialize(mp3_path, options=nil)
    raise ::ArgumentError.new('arg must be mp3 file') unless mp3_path =~ /\.mp3$/
    @mp3_path = mp3_path
    @options = options || {}
  end

  def process
    TempDir.create do |dir|
      Scissor(@mp3_path)[start, length] > "#{dir}/solamimi.mp3"
      mp3_to_wav("#{dir}/solamimi.mp3", "#{dir}/solamimi.wav")
      @japanese = text_from_wav("#{dir}/solamimi.wav")
    end
    self
  end

  def mp3_to_wav(source, target)
    system "ffmpeg -y -i #{source} -ac 1 -ar 16000 #{target} 2>/dev/null"
  end

  def text_from_wav(source)
    result = `echo #{source} | julius #{julius_options} 2>/dev/null`
    result.match(/\nsentence1: (.+?)\n/)[1].gsub(/\s/, '')
  end

  def julius_options
    "-charconv EUC-JP UTF-8 -C #{@options[:jconf_path]}"
  end

  def start
    @start ||= @options[:start] || 0
  end

  def length
    @length ||= @options[:length] || 5
  end
end

mp3_path, start, length = ARGV
jconf_path = Pit.get('julius', :require => {'conf_path' => 'your julius conf file path'})['conf_path']

start_info = start.split(':')
if start_info.size == 2
  sec_start = start_info[0].to_f * 60 + start_info[1].to_f
elsif start_info.size == 1
  sec_start = start.to_f
else
  exit 1
end
length = length.to_f || 5

s = Solamimi.listen_to(mp3_path, {
    :start => sec_start,
    :length => length,
    :jconf_path => jconf_path
})
puts "「#{s.japanese}」(#{start} - #{length})"