diarize-ruby 0.3.9 → 0.3.10

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 81c40e31baa740bf18b34d2de82bab3ae7897461
-  data.tar.gz: 3e987f962651badd8fc8a79d73a718d4660fe532
+  metadata.gz: 6538efc1cab1c18e656dcc50c1e2b57d8f9f38a7
+  data.tar.gz: d86715c3e757994d859c9847e013c5501895a5f1
 SHA512:
-  metadata.gz: c96801c0aa7dbf3409853fd519c18ce68ee291cb294f509ec642059152ac0d8cfc81936afe1b4c21d3ff4440b83e37c09594e36dddbe521e251697715653839f
-  data.tar.gz: a554a9ffc2de0f8f48aa198b7e11d3db639c49efbd941cad920f10453dc2c875ddd776a24c98685a14a3cac24f5890944efb947e1c143c1a8c81e63fada33587
+  metadata.gz: 0762b0b955406e9c43b95e44331f518379c90a3403addcb8478bdae93746095ed060805f4e128bfd3a7640118ba821b506e467d99b3dfe3679922056eeaafac4
+  data.tar.gz: 0a38dcab217b7b45595d734434be3ffbdf9755c7d869bb93a49198b288b45b59d2dcdb078213d2626fc67d70bb7d955b0eeaf5982a7cb3482ea647aa9af5328d

data/lib/diarize/speaker.rb

@@ -159,8 +159,7 @@ module Diarize
         'ws:gender' => gender,
         'ws:model' => model_uri,
         'ws:mean_log_likelihood' => mean_log_likelihood,
-        'ws:supervector_hash' => supervector.hash.to_s,
-        'ws:sha' => supervector.sha
+        'ws:supervector_hash' => supervector.hash.to_s
       }
     end
 
@@ -169,8 +168,7 @@ module Diarize
         'gender' => gender,
         'model' => model_uri,
         'mean_log_likelihood' => mean_log_likelihood,
-        'supervector_hash' => supervector.hash.to_s,
-        'sha' => supervector.sha
+        'supervector_hash' => supervector.hash.to_s
       }
     end
 

data/lib/diarize/super_vector.rb

@@ -3,65 +3,66 @@ module Diarize
     attr_reader :vector
 
     def initialize(vector)
-      @vector = vector
+      @vector = vector.is_a?(Array) ? Vector.elements(vector) : vector
     end
 
-    def self.generate_from_model(model)
-      # Generates a supervector from a LIUM GMM
-      dim = model.nb_of_components * model.components.get(0).dim
-      # vector = DoubleMatrix.new(1, dim)
-      # vector = Vector.elements(Array.new(dim, 0))
-      vector = Array.new(dim, 0)
-      model.nb_of_components.times do |k|
-        gaussian = model.components.get(k)
-        gaussian.dim.times do |i|
-          vector[k * gaussian.dim + i] = gaussian.mean(i)
+    class << self
+
+      def generate_from_model(model)
+        # Generates a supervector from a LIUM GMM
+        dim = model.nb_of_components * model.components.get(0).dim
+        vector = Array.new(dim, 0)
+        model.nb_of_components.times do |k|
+          gaussian = model.components.get(k)
+          gaussian.dim.times do |i|
+            vector[k * gaussian.dim + i] = gaussian.mean(i)
+          end
         end
+        SuperVector.new(Vector.elements(vector))
       end
-      # SuperVector.new(vector)
-      SuperVector.new(Vector.elements(vector))
-    end
 
-    def self.ubm_gaussian_weights
-      # Returns a vector of gaussian weights, same dimension as speaker's super vectors
-      @@ubm_gaussian_weights ||= begin
-        ubm = Speaker.ubm
-        # weights = DoubleMatrix.new(1, ubm.supervector.dim)
-        weights = Array.new(ubm.supervector.dim, 0)
-        ubm.model.nb_of_components.times do |k|
-          gaussian = ubm.model.components.get(k)
-          gaussian.dim.times do |i|
-            weights[k * gaussian.dim + i] = gaussian.weight
+      def ubm_gaussian_weights
+        # Returns a vector of gaussian weights, same dimension as speaker's super vectors
+        @@ubm_gaussian_weights ||= begin
+          ubm = Speaker.ubm
+          # weights = DoubleMatrix.new(1, ubm.supervector.dim)
+          weights = Array.new(ubm.supervector.dim, 0)
+          ubm.model.nb_of_components.times do |k|
+            gaussian = ubm.model.components.get(k)
+            gaussian.dim.times do |i|
+              weights[k * gaussian.dim + i] = gaussian.weight
+            end
           end
+          Vector.elements(weights)
         end
-        Vector.elements(weights)
       end
-    end
 
-    def self.ubm_covariance
-      # Returns a vector of diagonal covariances, same dimension as speaker's super vectors
-      @@ubm_covariance ||= begin
-        ubm = Speaker.ubm
-        # cov = DoubleMatrix.new(1, ubm.supervector.dim)
-        cov = Array.new(ubm.supervector.dim)
-        ubm.model.nb_of_components.times do |k|
-          gaussian = ubm.model.components.get(k)
-          gaussian.dim.times do |i|
-            cov[k * gaussian.dim + i] = gaussian.getCovariance(i, i)
+      def ubm_covariance
+        # Returns a vector of diagonal covariances, same dimension as speaker's super vectors
+        @@ubm_covariance ||= begin
+          ubm = Speaker.ubm
+          # cov = DoubleMatrix.new(1, ubm.supervector.dim)
+          cov = Array.new(ubm.supervector.dim)
+          ubm.model.nb_of_components.times do |k|
+            gaussian = ubm.model.components.get(k)
+            gaussian.dim.times do |i|
+              cov[k * gaussian.dim + i] = gaussian.getCovariance(i, i)
+            end
           end
+          Vector.elements(cov)
         end
-        Vector.elements(cov)
       end
-    end
 
-    def self.divergence(sv1, sv2)
-      # ubm_gaussian_weights.mul(((sv1.vector - sv2.vector) ** 2) / ubm_covariance).sum
-      diff   = sv1.vector - sv2.vector
-      square = diff.map {|el| el ** 2}
-      codiv  = Vector.elements(square.each.with_index.inject([]) {|a,(el,ix)| a << el / ubm_covariance[ix]})
-      mult   = ubm_gaussian_weights.each.with_index.inject([]) {|a,(el,ix)| a << el * codiv[ix]}
-      mult.inject(0, :+)
-    end
+      def divergence(sv1, sv2)
+        # ubm_gaussian_weights.mul(((sv1.vector - sv2.vector) ** 2) / ubm_covariance).sum
+        diff   = sv1.vector - sv2.vector
+        square = diff.map {|el| el ** 2}
+        codiv  = Vector.elements(square.each.with_index.inject([]) {|a,(el,ix)| a << el / ubm_covariance[ix]})
+        mult   = ubm_gaussian_weights.each.with_index.inject([]) {|a,(el,ix)| a << el * codiv[ix]}
+        mult.inject(0, :+)
+      end
+
+    end # class
 
     def dim
       @vector.size
@@ -71,10 +72,6 @@ module Diarize
       @vector.hash
     end
 
-    def sha
-      Digest::SHA256.hexdigest(hash.to_s)
-    end
-
     def to_a
       @vector.to_a
     end

data/lib/diarize/version.rb

@@ -1,3 +1,3 @@
 module Diarize
-  VERSION = "0.3.9"
+  VERSION = "0.3.10"
 end

data/test/speaker_test.rb

@@ -106,7 +106,6 @@ class SpeakerTest < Test::Unit::TestCase
     assert_equal true, to_rdf.include?("ws:model")
     assert_equal true, to_rdf.include?("ws:mean_log_likelihood")
     assert_equal true, to_rdf.include?("ws:supervector_hash")
-    assert_equal true, to_rdf.include?("ws:sha")
     assert_equal true, to_rdf.include?("https://www.example.com/model/1")
   end
 
@@ -118,7 +117,6 @@ class SpeakerTest < Test::Unit::TestCase
     assert_equal true, as_json.has_key?('model')
     assert_equal true, as_json.has_key?('mean_log_likelihood')
     assert_equal true, as_json.has_key?('supervector_hash')
-    assert_equal true, as_json.has_key?('sha')
   end
 
   def test_to_json
@@ -130,6 +128,5 @@ class SpeakerTest < Test::Unit::TestCase
     assert_equal true, as_json.has_key?('model')
     assert_equal true, as_json.has_key?('mean_log_likelihood')
     assert_equal true, as_json.has_key?('supervector_hash')
-    assert_equal true, as_json.has_key?('sha')
   end
 end

data/test/super_vector_test.rb

@@ -15,16 +15,24 @@ class SuperVectorTest < Test::Unit::TestCase
     end
   end
 
-  def test_hash
+  def test_initialize_from_supervector
     model = Diarize::Speaker.load_model(File.join(File.dirname(__FILE__), 'data', 'speaker1.gmm'))
-    sv = Diarize::SuperVector.generate_from_model(model)
-    assert_equal sv.vector.hash, sv.hash
+    sv1 = Diarize::SuperVector.generate_from_model(model)
+    sv2 = Diarize::SuperVector.new(sv1)
+    assert_equal sv1.hash, sv2.hash
+  end
+
+  def test_initialize_from_array
+    model = Diarize::Speaker.load_model(File.join(File.dirname(__FILE__), 'data', 'speaker1.gmm'))
+    sv1 = Diarize::SuperVector.generate_from_model(model)
+    sv2 = Diarize::SuperVector.new(sv1.to_a)
+    assert_equal sv1.hash, sv2.hash
   end
 
-  def test_sha
+  def test_hash
     model = Diarize::Speaker.load_model(File.join(File.dirname(__FILE__), 'data', 'speaker1.gmm'))
     sv = Diarize::SuperVector.generate_from_model(model)
-    assert_equal Digest::SHA256.hexdigest(sv.hash.to_s), sv.sha
+    assert_equal sv.vector.hash, sv.hash
   end
 
   def test_to_a

data/test/version_test.rb

@@ -3,7 +3,7 @@ require 'test_helper'
 class VersionTest < Test::Unit::TestCase
 
   def test_current_version
-    assert_equal "0.3.9", Diarize::VERSION
+    assert_equal "0.3.10", Diarize::VERSION
   end
 
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: diarize-ruby
 version: !ruby/object:Gem::Version
-  version: 0.3.9
+  version: 0.3.10
 platform: ruby
 authors:
 - Yves Raimond