pocketsphinx-ruby 0.2.0 → 0.3.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 3b38998dd59c577300db4b63bfda52e037c49409
-  data.tar.gz: 82dd10ee0f739d6459013513ae986d6e60124dfd
+  metadata.gz: 1be9c28022d45172119190075e2c4f340f129022
+  data.tar.gz: f00d504978967e201a795874fcdee1038cb96d03
 SHA512:
-  metadata.gz: d11e77f70106d9beaab587e7c8372dc217a9e511e7058e6d4f0a55fb4ce040b5ad50ef203a9755324ebaccc6fe919919eb04e5bb4770a290413b9990cecbfec6
-  data.tar.gz: fbea33173c9c8bdca4458c2dcb1f88e5643be169d8a54f4ed3baf593677f790d754e0f58a1c37ca527cbfeeeb9c0b19b28461afe417da509c2073543c33522ee
+  metadata.gz: 3407c5ce61d32c3576ce1fa6d16bc3249e0f74fc990b7b5dc8005c545e108cdca776e52e985354636a96ad626db2de90f71c5dd499ab29171b7b25a314cd5c26
+  data.tar.gz: b8be240f48b2962dd417c19589be7dba6a5b5460c9046deb66972507fe4655718884ad60d86ab8838c3428c946913e2e38ba5524bc1403f320ad01ee43a8e7f2

data/.travis.yml

@@ -1,10 +1,11 @@
 language: ruby
 rvm:
-  - 2.1.2
+  - 2.2.2
+  - 2.1.6
   - 2.0.0
   - 1.9.3
   - rbx-2.2.9
-  - jruby-1.7.16
+  - jruby-1.7.19
 before_install:
   - sudo apt-get update -qq
   - sudo apt-get install -y swig

data/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+**v0.3.0 - 17/04/15**
+
+* Add Words support to Decoder
+* Don't lazy initialize Pocketsphinx decoder
+* Use buffer size of 2048 as now [required](https://github.com/cmusphinx/pocketsphinx/commit/541b5dfa87ef5fffe509d7c195803fd45749db5e) by Pocketsphinx
+
+
 **v0.2.0 - 03/03/15**
 
 * Updated to latest Pocketsphinx API interface

data/README.md

@@ -69,7 +69,7 @@ The `AudioFileSpeechRecognizer` decodes directly from an audio file by coordinat
 recognizer = Pocketsphinx::AudioFileSpeechRecognizer.new
 
 recognizer.recognize('spec/assets/audio/goforward.raw') do |speech|
-  puts speech # => "go forward ten years"
+  puts speech # => "go forward ten meters"
 end
 ```
 
@@ -110,9 +110,9 @@ microphone = Pocketsphinx::Microphone.new
 
 File.open("test.raw", "wb") do |file|
   microphone.record do
-    FFI::MemoryPointer.new(:int16, 4096) do |buffer|
+    FFI::MemoryPointer.new(:int16, 2048) do |buffer|
       50.times do
-        sample_count = microphone.read_audio(buffer, 4096)
+        sample_count = microphone.read_audio(buffer, 2048)
         file.write buffer.get_bytes(0, sample_count * 2)
 
         sleep 0.1
@@ -133,7 +133,27 @@ The `Decoder` class uses Pocketsphinx's libpocketsphinx to decode audio data int
 decoder = Pocketsphinx::Decoder.new(Pocketsphinx::Configuration.default)
 decoder.decode 'spec/assets/audio/goforward.raw'
 
-puts decoder.hypothesis # => "go forward ten years"
+puts decoder.hypothesis # => "go forward ten meters"
+```
+
+And split into individual words with frame data:
+
+```ruby
+decoder.words
+# => [
+#  #<struct Pocketsphinx::Decoder::Word word="<s>", start_frame=608, end_frame=610>,
+#  #<struct Pocketsphinx::Decoder::Word word="go", start_frame=611, end_frame=622>,
+#  #<struct Pocketsphinx::Decoder::Word word="forward", start_frame=623, end_frame=675>,
+#  #<struct Pocketsphinx::Decoder::Word word="ten", start_frame=676, end_frame=711>,
+#  #<struct Pocketsphinx::Decoder::Word word="meters", start_frame=712, end_frame=770>,
+#  #<struct Pocketsphinx::Decoder::Word word="</s>", start_frame=771, end_frame=821>
+# ]
+```
+
+Note: When the `Decoder` is initialized, the supplied `Configuration` is updated by Pocketsphinx with some settings from the acoustic model. To see exactly what's going on:
+
+```ruby
+Pocketsphinx::Decoder.new(Pocketsphinx::Configuration.default).configuration.changes
 ```
 
 

data/examples/decode_audio_file.rb

@@ -8,4 +8,4 @@ include Pocketsphinx
 decoder = Decoder.new(Configuration.default)
 decoder.decode 'spec/assets/audio/goforward.raw'
 
-puts decoder.hypothesis # => "go forward ten years"
+puts decoder.hypothesis # => "go forward ten meters"

data/examples/record_audio_file.rb

@@ -5,7 +5,7 @@ require "pocketsphinx-ruby"
 
 include Pocketsphinx
 
-MAX_SAMPLES = 4096
+MAX_SAMPLES = 2048
 RECORDING_INTERVAL = 0.1
 RECORDING_LENGTH = 5
 

data/lib/pocketsphinx/api/pocketsphinx.rb

@@ -7,7 +7,12 @@ module Pocketsphinx
       typedef :pointer, :decoder
       typedef :pointer, :configuration
 
-      attach_function :ps_init, [:configuration], :decoder
+      # Allows expect(API::Pocketsphinx).to receive(:ps_init) in JRuby specs
+      def self.ps_init(*args)
+        ps_init_private(*args)
+      end
+
+      attach_function :ps_init_private, :ps_init, [:configuration], :decoder
       attach_function :ps_reinit, [:decoder, :configuration], :int
       attach_function :ps_default_search_args, [:pointer], :void
       attach_function :ps_args, [], :pointer
@@ -21,6 +26,15 @@ module Pocketsphinx
       attach_function :ps_unset_search, [:decoder, :string], :int
       attach_function :ps_get_search, [:decoder], :string
       attach_function :ps_set_search, [:decoder, :string], :int
+
+      typedef :pointer, :seg_iter
+
+      attach_function :ps_seg_iter, [:decoder, :pointer], :seg_iter
+      attach_function :ps_seg_next, [:seg_iter], :seg_iter
+      attach_function :ps_seg_word, [:seg_iter], :string
+      attach_function :ps_seg_frames, [:seg_iter, :pointer, :pointer], :void
+      attach_function :ps_seg_prob, [:seg_iter, :pointer, :pointer, :pointer], :int32
+      attach_function :ps_seg_free, [:seg_iter], :void
     end
   end
 end

data/lib/pocketsphinx/audio_file.rb

@@ -6,7 +6,7 @@ module Pocketsphinx
     # @param [FFI::Pointer] buffer 16bit buffer of at least max_samples in size
     # @params [Fixnum] max_samples The maximum number of samples to read from the audio file
     # @return [Fixnum] Samples actually read; nil if EOF
-    def read_audio(buffer, max_samples = 4096)
+    def read_audio(buffer, max_samples = 2048)
       if file.nil?
         raise "Can't read audio: use AudioFile#start_recording to open the file first"
       end

data/lib/pocketsphinx/audio_file_speech_recognizer.rb

@@ -1,7 +1,7 @@
 module Pocketsphinx
   # High-level class for live speech recognition from a raw audio file.
   class AudioFileSpeechRecognizer < SpeechRecognizer
-    def recognize(file_path, max_samples = 4096)
+    def recognize(file_path, max_samples = 2048)
       self.recordable = AudioFile.new(file_path)
 
       super(max_samples) do |speech|

data/lib/pocketsphinx/decoder.rb

@@ -1,5 +1,5 @@
 module Pocketsphinx
-  class Decoder < Struct.new(:configuration)
+  class Decoder
     require 'delegate'
 
     include API::CallHelpers
@@ -13,7 +13,22 @@ module Pocketsphinx
       end
     end
 
+    Word = Struct.new(:word, :start_frame, :end_frame)
+
     attr_writer :ps_api
+    attr_accessor :configuration
+
+    # Initialize a Decoder
+    #
+    # Note that this initialization process actually updates the Configuration based on settings
+    # which are found in feat.params along with the acoustic model.
+    #
+    # @param [Configuration] configuration
+    # @param [FFI::Pointer] ps_decoder An optional Pocketsphinx decoder. One is initialized if not provided.
+    def initialize(configuration, ps_decoder = nil)
+      @configuration = configuration
+      init_decoder if ps_decoder.nil?
+    end
 
     # Reinitialize the decoder with updated configuration.
     #
@@ -108,6 +123,30 @@ module Pocketsphinx
       )
     end
 
+    # Get an array of words with start/end frame values (10msec/frame) for current hypothesis
+    #
+    # @return [Array] Array of words with start/end frame values (10msec/frame)
+    def words
+      mp_path_score = FFI::MemoryPointer.new(:int32, 1)
+      start_frame   = FFI::MemoryPointer.new(:int32, 1)
+      end_frame     = FFI::MemoryPointer.new(:int32, 1)
+
+      seg_iter = ps_api.ps_seg_iter(ps_decoder, mp_path_score)
+      words    = []
+
+      until seg_iter.null? do
+        ps_api.ps_seg_frames(seg_iter, start_frame, end_frame)
+        words << Pocketsphinx::Decoder::Word.new(
+          ps_api.ps_seg_word(seg_iter),
+          start_frame.get_int32(0),
+          end_frame.get_int32(0)
+        )
+        seg_iter = ps_api.ps_seg_next(seg_iter)
+      end
+
+      words
+    end
+
     # Adds new search using JSGF model.
     #
     # Convenience method to parse JSGF model from string and create a search.

data/lib/pocketsphinx/microphone.rb

@@ -49,7 +49,7 @@ module Pocketsphinx
     # @params [Fixnum] max_samples The maximum number of samples to read from the audio device
     # @return [Fixnum] Samples actually read (could be 0 since non-blocking); nil if not
     #   recording and no more samples remaining to be read from most recent recording.
-    def read_audio(buffer, max_samples = 4096)
+    def read_audio(buffer, max_samples = 2048)
       samples = ps_api.ad_read(@ps_audio_device, buffer, max_samples)
       samples if samples >= 0
     end
@@ -60,7 +60,7 @@ module Pocketsphinx
     # we specify a delay which should fill half of the max buffer size
     #
     # @param [Fixnum] max_samples The maximum samples we tried to read from the audio device
-    def read_audio_delay(max_samples = 4096)
+    def read_audio_delay(max_samples = 2048)
       max_samples.to_f / (2 * sample_rate)
     end
 

data/lib/pocketsphinx/speech_recognizer.rb

@@ -43,7 +43,7 @@ module Pocketsphinx
     # Recognize speech and yield hypotheses in infinite loop
     #
     # @param [Fixnum] max_samples Number of samples to process at a time
-    def recognize(max_samples = 4096, &b)
+    def recognize(max_samples = 2048, &b)
       unless ALGORITHMS.include?(algorithm)
         raise NotImplementedError, "Unknown speech recognition algorithm: #{algorithm}"
       end

data/lib/pocketsphinx/version.rb

@@ -1,3 +1,3 @@
 module Pocketsphinx
-  VERSION = "0.2.0"
+  VERSION = "0.3.0"
 end

data/spec/configuration_spec.rb

@@ -65,7 +65,7 @@ describe Pocketsphinx::Configuration do
 
   describe '#setting_names' do
     it 'contains the names of all possible system settings' do
-      expect(subject.setting_names.count).to eq(112)
+      expect(subject.setting_names.count).to eq(114)
     end
   end
 
@@ -84,7 +84,7 @@ describe Pocketsphinx::Configuration do
     it 'gives details for all settings when no name is specified' do
       details = subject.details
 
-      expect(details.count).to eq(112)
+      expect(details.count).to eq(114)
       expect(details.first).to eq({
         name: "agc",
         type: :string,

data/spec/decoder_spec.rb

@@ -1,7 +1,7 @@
 require 'spec_helper'
 
 describe Pocketsphinx::Decoder do
-  subject { Pocketsphinx::Decoder.new(configuration) }
+  subject { Pocketsphinx::Decoder.new(configuration, ps_decoder) }
   let(:ps_api) { subject.ps_api }
   let(:ps_decoder) { double }
   let(:configuration) { Pocketsphinx::Configuration.default }
@@ -11,6 +11,17 @@ describe Pocketsphinx::Decoder do
     allow(ps_api).to receive(:ps_init).and_return(ps_decoder)
   end
 
+  describe 'initialization' do
+    it 'initializes the underlying Pocketsphinx decoder when one is not provided' do
+      expect(Pocketsphinx::API::Pocketsphinx)
+        .to receive(:ps_init)
+        .with(configuration.ps_config)
+        .and_return(ps_decoder)
+
+      Pocketsphinx::Decoder.new(configuration)
+    end
+  end
+
   describe '#reconfigure' do
     it 'calls libpocketsphinx and the configuration post initialize hook' do
       expect(ps_api)
@@ -53,24 +64,24 @@ describe Pocketsphinx::Decoder do
 
   describe '#process_raw' do
     it 'calls libpocketsphinx' do
-      FFI::MemoryPointer.new(:int16, 4096) do |buffer|
+      FFI::MemoryPointer.new(:int16, 2048) do |buffer|
         expect(ps_api)
           .to receive(:ps_process_raw)
-          .with(subject.ps_decoder, buffer, 4096, 0, 0)
+          .with(subject.ps_decoder, buffer, 2048, 0, 0)
           .and_return(0)
 
-        subject.process_raw(buffer, 4096, false, false)
+        subject.process_raw(buffer, 2048, false, false)
       end
     end
 
     it 'raises an exception on error' do
-      FFI::MemoryPointer.new(:int16, 4096) do |buffer|
+      FFI::MemoryPointer.new(:int16, 2048) do |buffer|
         expect(ps_api)
           .to receive(:ps_process_raw)
-          .with(subject.ps_decoder, buffer, 4096, 0, 0)
+          .with(subject.ps_decoder, buffer, 2048, 0, 0)
           .and_return(-1)
 
-        expect { subject.process_raw(buffer, 4096, false, false) }
+        expect { subject.process_raw(buffer, 2048, false, false) }
           .to raise_exception "Decoder#process_raw failed with error code -1"
       end
     end
@@ -148,6 +159,35 @@ describe Pocketsphinx::Decoder do
     end
   end
 
+  describe '#words' do
+    let(:iterator) { FFI::MemoryPointer.from_string("") }
+
+    it 'calls libpocketsphinx' do
+      expect(ps_api).to receive(:ps_seg_iter).ordered.and_return(iterator)
+
+      expect(ps_api).to receive(:ps_seg_frames).ordered do |seg_iter, start_frame, end_frame|
+        start_frame.put_int16(0, 10)
+        end_frame.put_int16(0, 20)
+      end
+
+      expect(ps_api).to receive(:ps_seg_word).ordered.and_return("one")
+      expect(ps_api).to receive(:ps_seg_next).ordered.and_return(iterator)
+
+      expect(ps_api).to receive(:ps_seg_frames).ordered do |seg_iter, start_frame, end_frame|
+        start_frame.put_int16(0, 30)
+        end_frame.put_int16(0, 40)
+      end
+
+      expect(ps_api).to receive(:ps_seg_word).ordered.and_return("two")
+      expect(ps_api).to receive(:ps_seg_next).ordered.and_return(FFI::Pointer::NULL)
+
+      words = subject.words
+
+      expect(words[0]).to eq(Pocketsphinx::Decoder::Word.new("one", 10, 20))
+      expect(words[1]).to eq(Pocketsphinx::Decoder::Word.new("two", 30, 40))
+    end
+  end
+
   describe '#set_jsgf_string' do
     it 'calls libpocketsphinx' do
       expect(ps_api)

data/spec/integration/decoder_spec.rb

@@ -5,22 +5,44 @@ describe Pocketsphinx::Decoder do
   let(:configuration) { @configuration }
 
   # Share decoder across all examples for speed
-  before :all do
+  before do
     @configuration = Pocketsphinx::Configuration.default
     @decoder = Pocketsphinx::Decoder.new(@configuration)
   end
 
+  it 'reads cmninit configuration values from default acoustic model feat.params' do
+    expect(configuration.details('cmninit')[:default]).to eq("8.0")
+    expect(configuration.details('cmninit')[:value]).to eq("40,3,-1")
+  end
+
   describe '#decode' do
     it 'correctly decodes the speech in goforward.raw' do
-      @decoder.ps_api = nil
       subject.decode File.open('spec/assets/audio/goforward.raw', 'rb')
-
       expect(subject.hypothesis).to eq("go forward ten meters")
     end
 
+    # FIXME: This test illustrates a current issue discussed in:
+    #        https://github.com/watsonbox/pocketsphinx-ruby/issues/10
+    it 'incorrectly decodes the speech in hello.wav on first attempt' do
+      hypotheses = (1..2).map do
+        subject.decode File.open('spec/assets/audio/hello.wav', 'rb')
+        subject.hypothesis
+      end
+
+      expect(hypotheses).to eq(['oh', 'hello'])
+    end
+
     it 'accepts a file path as well as a stream' do
       subject.decode 'spec/assets/audio/goforward.raw'
       expect(subject.hypothesis).to eq("go forward ten meters")
     end
+
+    it 'reports words with start/end frame values' do
+      subject.decode File.open('spec/assets/audio/goforward.raw', 'rb')
+
+      expect(subject.words.map(&:word)).to eq(["<s>", "go", "forward", "ten", "meters", "</s>"])
+      expect(subject.words.map(&:start_frame)).to eq([0, 46, 64, 117, 153, 212])
+      expect(subject.words.map(&:end_frame)).to eq([45, 63, 116, 152, 211, 260])
+    end
   end
 end

data/spec/integration/default_recognition_spec.rb

@@ -14,7 +14,7 @@ describe 'speech recognition with default configuration' do
 
   describe '#recognize' do
     it 'should decode speech in raw audio' do
-      expect { |b| subject.recognize('spec/assets/audio/goforward.raw', 4096, &b) }.
+      expect { |b| subject.recognize('spec/assets/audio/goforward.raw', 2048, &b) }.
         to yield_with_args("go forward ten meters")
     end
   end

data/spec/integration/grammar_recognition_spec.rb

@@ -18,7 +18,7 @@ describe 'speech recognition with a grammar' do
 
   describe '#recognize' do
     it 'should decode speech in raw audio' do
-      expect { |b| subject.recognize(4096, &b) }.to yield_with_args("go forward ten meters")
+      expect { |b| subject.recognize(2048, &b) }.to yield_with_args("go forward ten meters")
     end
   end
 end

data/spec/integration/keyword_recognition_spec.rb

@@ -19,7 +19,7 @@ describe 'keyword spotting' do
 
   describe '#recognize' do
     it 'should decode speech in raw audio' do
-      expect { |b| subject.recognize(4096, &b) }.to yield_with_args('forward')
+      expect { |b| subject.recognize(2048, &b) }.to yield_with_args('forward')
     end
   end
 end

data/spec/microphone_spec.rb

@@ -72,16 +72,16 @@ describe Pocketsphinx::Microphone do
     it 'calls libsphinxad' do
       expect(ps_api)
         .to receive(:ad_read)
-        .with(subject.ps_audio_device, :buffer, 4096)
+        .with(subject.ps_audio_device, :buffer, 2048)
         .and_return(0)
 
-      subject.read_audio(:buffer, 4096)
+      subject.read_audio(:buffer, 2048)
     end
   end
 
   describe '#read_audio_delay' do
-    it 'should be 0.128 seconds for a max_samples of 4096 and sample rate of 16kHz' do
-      expect(subject.read_audio_delay(4096)).to eq(0.128)
+    it 'should be 0.064 seconds for a max_samples of 2048 and sample rate of 16kHz' do
+      expect(subject.read_audio_delay(2048)).to eq(0.064)
     end
   end
 

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: pocketsphinx-ruby
 version: !ruby/object:Gem::Version
-  version: 0.2.0
+  version: 0.3.0
 platform: ruby
 authors:
 - Howard Wilson
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2015-03-03 00:00:00.000000000 Z
+date: 2015-04-17 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ffi
@@ -122,6 +122,7 @@ files:
 - lib/pocketsphinx/version.rb
 - pocketsphinx-ruby.gemspec
 - spec/assets/audio/goforward.raw
+- spec/assets/audio/hello.wav
 - spec/assets/grammars/goforward.gram
 - spec/assets/grammars/invalid.gram
 - spec/assets/grammars/sentences.gram
@@ -162,6 +163,7 @@ specification_version: 4
 summary: Ruby speech recognition with Pocketsphinx
 test_files:
 - spec/assets/audio/goforward.raw
+- spec/assets/audio/hello.wav
 - spec/assets/grammars/goforward.gram
 - spec/assets/grammars/invalid.gram
 - spec/assets/grammars/sentences.gram