pihsi 0.0.8 → 0.1.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 650af7f9bd97d82f5d9163c4572835755f7175bb
-  data.tar.gz: c1e853d14cd7ce5110b332357041a17cc839c0fc
+  metadata.gz: d949dd5f7e6d5f382c6c74a44f3678fbfe0a8fa8
+  data.tar.gz: ed8b316d1c8ffb11f954a57a0389d787079acd70
 SHA512:
-  metadata.gz: f0ce6a10c92c009980f9500a3cb6bdca5984e3671b433ae2862d59532acf94c575b05e4c1241bb2268291fe4a99a77db52d3d56e09a99e78cdda601ca2e8c2d8
-  data.tar.gz: 00df6ac4921be463b5bef64df63c18d563d4f6bc87d6162704b689aae22161fb3ede23895366e9067604b3e1e7ded91e39da23cfcb1c7ae449f4a7e95c771c0a
+  metadata.gz: 5b09454ee9df43e8aac32714dcaf0a9c11fa7a56f5ace89730cd817fdba4115a4cbca1199ce4ef04e30f3f82ae6727bb7ebf037be84a0d02a6fb3c8ec0a2c7b9
+  data.tar.gz: 31ccc59f39331163b8fbeb290e7ee2db0e2ded8e40f9b77a5046610cbbdb5f79d34d40b5cf3fdc660aca5842699896d6e43a2b7ba48db23cb9db293b6174a694

data/README.md

@@ -29,7 +29,7 @@ Or install it yourself as:
 Initialize Pihsi::SpeechRecognizer with proper [hmm, lm and dict](http://cmusphinx.sourceforge.net/wiki/tutorialpocketsphinx#initialization):
 
 ```ruby
-recognizer = Pihsi::SpeechRecognizer.new
+recognizer = Pihsi::SpeechRecognizer.new hmm: 'xxx', lm: 'yyy', dict: 'zzz'
 ```
 
 Recognize a string read from your audio file:

data/ext/pocket_sphinx/pocket_sphinx.c

@@ -5,50 +5,40 @@ typedef struct ps {
     ps_decoder_t *decoder;
 } PocketSphinx;
 
-VALUE decode(VALUE self, VALUE data) {
+static VALUE rb_eUtteranceError;
+
+/* Converts raw audio data into text.
+ *
+ * @param data [String] the raw audio data
+ * @return [String, nil] the transcribed text or nil
+ */
+VALUE recognize(VALUE self, VALUE data) {
     char const *hyp, *uttid;
     int rv;
     int32 score;
     ps_decoder_t *ps;
     PocketSphinx *pocketSphinx;
-    VALUE result = rb_hash_new();
 
     Data_Get_Struct(self, PocketSphinx, pocketSphinx);
     ps = pocketSphinx -> decoder;
 
-    int data_length = RARRAY_LEN(data);
-    int16 *c_data = malloc(sizeof(int16) * data_length);
-
-    int i;
-    for(i = 0; i < data_length; i++) {
-        c_data[i] = NUM2INT(rb_ary_entry(data, i)); // Rubinius does not support NUM2SHORT();
-    }
-
     rv = ps_start_utt(ps, "goforward");
 
-    if (rv < 0) {
-        free(c_data);
-        rb_raise(rb_eStandardError, "cannot start utterance");
-    }
+    if (rv < 0)
+        rb_raise(rb_eUtteranceError, "cannot start utterance");
 
-    rv = ps_process_raw(ps, c_data, data_length, FALSE, FALSE);
+    rv = ps_process_raw(ps, (int16 *) RSTRING_PTR(data), RSTRING_LEN(data) / 2, FALSE, FALSE);
     rv = ps_end_utt(ps);
 
-    if (rv < 0) {
-        free(c_data);
-        rb_raise(rb_eStandardError, "cannot end utterance");
-    }
+    if (rv < 0)
+        rb_raise(rb_eUtteranceError, "cannot end utterance");
 
     hyp = ps_get_hyp(ps, &score, &uttid);
 
-    free(c_data);
-
     if (hyp == NULL) {
         return Qnil;
     } else {
-        rb_hash_aset(result, rb_str_new2("score"), INT2NUM(hyp));
-        rb_hash_aset(result, rb_str_new2("hypothesis"), rb_str_new2(hyp));
-        return result;
+        return rb_str_new2(hyp);
     }
 }
 
@@ -63,16 +53,22 @@ static VALUE allocate(VALUE self) {
     return Data_Make_Struct(self, PocketSphinx, 0, deallocate, ps);
 }
 
-VALUE initialize(VALUE self, VALUE options) {
+/*
+ * @param options [Hash] PocketSphinx configuration (e.g. hmm)
+ */
+VALUE initialize(VALUE self, VALUE options_hash) {
     int i;
     PocketSphinx *ps;
     VALUE option, key, value;
     char *c_key;
     cmd_ln_t *config = cmd_ln_init(NULL, ps_args(), TRUE, "-logfn", "/dev/null", NULL);
 
+    VALUE options = rb_funcall(options_hash, rb_intern("to_a"), 0);
+
     for (i = 0; i < RARRAY_LEN(options); i++) {
         option = rb_ary_entry(options, i);
-        key = rb_ary_entry(option, 0);
+        key = rb_funcall(rb_ary_entry(option, 0), rb_intern("to_s"), 0);
+        key = rb_funcall(key, rb_intern("prepend"), 1, rb_str_new2("-"));
         c_key = RSTRING_PTR(key);
         value = rb_ary_entry(option, 1);
 
@@ -95,9 +91,9 @@ VALUE initialize(VALUE self, VALUE options) {
 
 void Init_pocket_sphinx() {
     VALUE rb_mPihsi = rb_define_module("Pihsi");
-    VALUE rb_mPocketSphinx = rb_define_module_under(rb_mPihsi, "PocketSphinx");
-    VALUE rb_cDecoder = rb_define_class_under(rb_mPocketSphinx, "Decoder", rb_cObject);
-    rb_define_alloc_func(rb_cDecoder, allocate);
-    rb_define_method(rb_cDecoder, "initialize", initialize, 1);
-    rb_define_method(rb_cDecoder, "decode", decode, 1);
+    VALUE rb_cSpeechRecognizer = rb_define_class_under(rb_mPihsi, "SpeechRecognizer", rb_cObject);
+    rb_define_alloc_func(rb_cSpeechRecognizer, allocate);
+    rb_define_method(rb_cSpeechRecognizer, "initialize", initialize, 1);
+    rb_define_method(rb_cSpeechRecognizer, "recognize", recognize, 1);
+    rb_eUtteranceError = rb_define_class_under(rb_mPihsi, "UtteranceError", rb_eStandardError);
 }

data/lib/pihsi/version.rb

@@ -1,3 +1,3 @@
 module Pihsi
-  VERSION = "0.0.8"
+  VERSION = "0.1.0"
 end

data/lib/pihsi.rb

@@ -1,7 +1,7 @@
 require 'rubygems'
 require 'bundler/setup'
 require "pihsi/version"
-require "pihsi/speech_recognizer"
+require "pocket_sphinx/pocket_sphinx"
 
 module Pihsi
   # Your code goes here...

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: pihsi
 version: !ruby/object:Gem::Version
-  version: 0.0.8
+  version: 0.1.0
 platform: ruby
 authors:
 - Zhi-Qiang Lei
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2014-09-04 00:00:00.000000000 Z
+date: 2014-09-07 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bundler
@@ -112,7 +112,6 @@ files:
 - ext/pocket_sphinx/extconf.rb
 - ext/pocket_sphinx/pocket_sphinx.c
 - lib/pihsi.rb
-- lib/pihsi/speech_recognizer.rb
 - lib/pihsi/version.rb
 - pihsi.gemspec
 - spec/factories/speech_recognizer.rb

data/lib/pihsi/speech_recognizer.rb

@@ -1,24 +0,0 @@
-require "pocket_sphinx/pocket_sphinx"
-
-module Pihsi
-  class SpeechRecognizer
-    attr_reader :decoder
-
-    def initialize(options = {})
-      _options = options.inject([]) do |result, (key, value)|
-        result << ["-#{key}", value] unless value.nil?
-        result
-      end
-      @decoder = PocketSphinx::Decoder.new(_options)
-    end
-
-    # Converts raw audio data into text
-    # @param data [String] the raw audio data
-    # @return [String, nil] the transcribed text or nil
-    def recognize(data)
-      if result = decoder.decode(data.unpack('s*'))
-        result["hypothesis"]
-      end
-    end
-  end
-end