aubio 0.2.0 → 0.2.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA1:
-  metadata.gz: 1b214156bf7a967baf0ab368060f8cef89ec8b3b
-  data.tar.gz: 3b2943f248a6c8d2657fa8640fe328742e185a1b
+  metadata.gz: 1acc7f11dab076c36e450cf7a1d4ec1c96aed13f
+  data.tar.gz: 8a47f61e2797d48ea428914328ba32e0f772b630
 SHA512:
-  metadata.gz: 9b6c9336bcb1a83d00cdde450e26e16026526d66c6fe3189c647cd98adbedd7197293942291fc92b68ae38750d42f87d4110d363626af2ed9e52cbb9ad944a41
-  data.tar.gz: 67e7947c7dac37ba5f30c9f6cc534f31ed46bdd9c46a82861b17ac052838b4d80723e1ca98e2c3127de8e0bf6edd8b40fa9ee1b1881b4982af613c85a547041d
+  metadata.gz: cf26ded45b899acf39ec1b82e6a373506df549e18f5acc441265e9d914943053988d09fb4c5ce648e184c18c99830ff0192fdbd8e2b9de72f99295dd850d0620
+  data.tar.gz: fb09d46d4a1e70f2a6a5d73b8e5a6f69934b8dc3fc9e9e3a555576c3826290ce853d776fcad1f962f0b48b582fcbc78b33b9a6a23e727b2a3907be267ba2afd3

data/README.md

@@ -42,8 +42,8 @@ From there you can access the following:
 
 ```
 my_file.onsets # list of extracted onsets
-# NOT YET IMPLEMENTED # my_file.pitches # list of extracted pitches
-# NOT YET IMPLEMENTED # my_file.beats # where one would tap their foot
+my_file.pitches # list of extracted pitches
+my_file.beats # where one would tap their foot
 # NOT YET IMPLEMENTED # my_file.notes # list of onsets with pitches
 # NOT YET IMPLEMENTED # my_file.silences # list of silent regions
 # NOT YET IMPLEMENTED # my_file.mel_frequency_cepstral_coefficients # list of mfccs
@@ -51,15 +51,25 @@ my_file.onsets # list of extracted onsets
 
 All of these are Ruby `Enumerator`s which means they'll respond to `.next`, and so on. If you want the whole list all at once call `.to_a` on it.
 
+### Avoiding memory leaks
+
+The underlying C library allocates some memory so to clean this up you'll need to run
+
+    my_file.close
+
+to free it up.
+
 ## Data format
 
 Each "event" that `aubio` describes is represented as a `Hash` like so:
 
 ```
-my_file.onsets.first #=> {s: 0.0, ms: 0.0}
+my_file.onsets.first #=> {s: 0.0, ms: 0.0, start: 1, end: 0}
 ```
 
-`s` and `ms` refer to seconds and milliseconds respectively. 
+`s` and `ms` refer to seconds and milliseconds respectively.
+
+`start: 1` and `end: 1` are special events that describe the start and the end of the audio file respectively. Whilst the `start` event at `0.0s` is usually an onset, the `end` is a convenience added by the Ruby wrapper to allow for easier slicing of sound files into samples.
 
 ### Still to implement
 
@@ -88,10 +98,8 @@ Aubio.open("/path/to/audio/file", sample_rate: 44100)
 ## Bugs / Still to do
 
 * better tests
-* add end of file as optional onset for slicing purposes
-* implement relative timing in offset
-* use `Offsets` class as a basis to implement the other functions
-* implement class level methods for "bpm"
+* use `Onsets` class as a basis to implement the other functions
+* improve accuracy of bpm - seems to be consistently too fast on things I've tried
 * look into streaming results for live inputs
 
 ## Development

data/lib/aubio.rb

@@ -1,11 +1,16 @@
 require_relative "aubio/version"
 require_relative "aubio/api"
 require_relative "aubio/onsets"
+require_relative "aubio/pitches"
+require_relative "aubio/beats"
 
 module Aubio
-  class Base
-    class FileNotFound < Exception; end
+  class AubioException < Exception; end
+  class FileNotFound < AubioException; end
+  class AlreadyClosed < AubioException; end
+  class InvalidAudioInput < AubioException; end
 
+  class Base
     def initialize(path, params)
       raise FileNotFound unless File.file?(path)
 
@@ -14,11 +19,71 @@ module Aubio
 
 			@source = Api.new_aubio_source(path, sample_rate, hop_size)
       @params = params
+
+      check_for_valid_audio_source(path)
+    end
+
+    def close
+      Api.del_aubio_source(@source)
+      @is_closed = true
     end
 
     def onsets
+      check_for_closed
+
       Onsets.new(@source, @params).each
     end
+
+    def pitches
+      check_for_closed
+
+      Pitches.new(@source, @params).each
+    end
+
+    def beats
+      check_for_closed
+
+      Beats.new(@source, @params).each
+    end
+
+    def bpm
+      check_for_closed
+
+      beat_locations = Beats.new(@source, @params).each.to_a
+      beat_periods = beat_locations.each_cons(2).map {|a,b| b[:s] - a[:s] }
+
+      # use interquartile median to discourage outliers
+      s = beat_periods.length
+      qrt_lower_idx = (s/4.0).floor
+      qrt_upper_idx = qrt_lower_idx * 3
+      interquartile_beat_periods = beat_periods[qrt_lower_idx..qrt_upper_idx]
+
+      # Calculate median
+      iqs = interquartile_beat_periods.length
+
+      # debug
+      interquartile_beat_periods.sort.each {|x| puts x }
+
+      iq_median_beat_period = interquartile_beat_periods.sort[(iqs/2.0).floor() - 1]
+      60.0 / iq_median_beat_period
+    end
+
+    private
+    def check_for_closed
+      raise AlreadyClosed if @is_closed
+    end
+
+    def check_for_valid_audio_source(path)
+      begin
+        @source.read_pointer
+      rescue FFI::NullPointerError
+        raise InvalidAudioInput.new(%Q{
+
+            Couldn't read file at #{path}
+            Did you install aubio with libsndfile support?
+          })
+      end
+    end
   end
 end
 

data/lib/aubio/beats.rb

@@ -0,0 +1,75 @@
+module Aubio
+	class Beats
+
+		def initialize(aubio_source, params)
+      # TODO: cleanup param dups
+			@sample_rate = params[:sample_rate] || 44100
+			@window_size = params[:window_size] || 1024
+			@hop_size    = params[:hop_size]    || 512
+
+			@source = aubio_source
+			@tempo = Api.new_aubio_tempo('specdiff', @window_size, @hop_size, @sample_rate)
+
+			# create output for source
+			@sample_buffer = Api.new_fvec(@hop_size)
+			# create output for beat
+			@out_fvec = Api.new_fvec(1)
+		end
+
+		def each
+			return enum_for(:each) unless block_given?
+
+			total_frames_counter = 0
+			read_buffer = FFI::MemoryPointer.new(:int)
+
+			loop do
+				# Perform tempo calculation
+        Api.aubio_source_do(@source, @sample_buffer, read_buffer)
+				Api.aubio_tempo_do(@tempo, @sample_buffer, @out_fvec)
+
+        # Retrieve result
+				is_beat = Api.fvec_get_sample(@out_fvec, 0)
+        no_of_bytes_read = read_buffer.read_int
+        total_frames_counter += no_of_bytes_read
+
+        if is_beat > 0.0
+					tempo_seconds = Api.aubio_tempo_get_last_s(@tempo)
+					tempo_milliseconds = Api.aubio_tempo_get_last_ms(@tempo)
+          tempo_confidence = Api.aubio_tempo_get_confidence(@tempo)
+
+					output = {
+            :confidence => tempo_confidence,
+            :s => tempo_seconds,
+            :ms => tempo_milliseconds,
+            :start => (tempo_seconds == 0.0 ? 1 : 0),
+            :end => 0
+					}
+          yield output
+				end
+
+        if no_of_bytes_read != @hop_size
+          # there's no more audio to look at
+
+          # Let's output one last tempo to mark the end of the file
+          total_time = total_frames_counter.to_f / @sample_rate.to_f
+          output = {
+            :confidence => 1.0,
+            :s => total_time,
+            :ms => total_time/1000.0,
+            :start => 0,
+            :end  => 1
+          }
+          yield output
+
+          # clean up
+          Api.del_aubio_tempo(@tempo)
+          Api.del_fvec(@sample_buffer)
+          Api.del_fvec(@out_fvec)
+
+          break
+        end
+			end
+		end
+
+	end
+end

data/lib/aubio/onsets.rb

@@ -22,37 +22,50 @@ module Aubio
 			return enum_for(:each) unless block_given?
 
 			total_frames_counter = 0
-			tmp_read = FFI::MemoryPointer.new(:int)
+			read_buffer = FFI::MemoryPointer.new(:int)
 
 			loop do
 				# Perform onset calculation
-        Api.aubio_source_do(@source, @sample_buffer, tmp_read)
+        Api.aubio_source_do(@source, @sample_buffer, read_buffer)
 				Api.aubio_onset_do(@onset, @sample_buffer, @out_fvec)
 
         # Retrieve result
 				onset_new_peak = Api.fvec_get_sample(@out_fvec, 0)
+        no_of_bytes_read = read_buffer.read_int
+        total_frames_counter += no_of_bytes_read
 
         if onset_new_peak > 0.0
 					onset_seconds = Api.aubio_onset_get_last_s(@onset)
 					onset_milliseconds = Api.aubio_onset_get_last_ms(@onset)
-          # TODO: implement relative here
 					output = {
             :s => onset_seconds,
-            :ms => onset_milliseconds
+            :ms => onset_milliseconds,
+            :start => (onset_seconds == 0.0 ? 1 : 0),
+            :end => 0
 					}
           yield output
 				end
 
-				read = tmp_read.read_int
-				total_frames_counter += read
-				if(read != @hop_size) then
+        if no_of_bytes_read != @hop_size
+          # there's no more audio to look at
+
+          # Let's output one last onset to mark the end of the file
+          total_time = total_frames_counter.to_f / @sample_rate.to_f
+          output = {
+            :s => total_time,
+            :ms => total_time/1000.0,
+            :start => 0,
+            :end  => 1
+          }
+          yield output
+
           # clean up
-          Api.del_aubio_source(@source)
-					Api.del_fvec(@sample_buffer)
-					Api.del_fvec(@out_fvec)
+          Api.del_aubio_onset(@onset)
+          Api.del_fvec(@sample_buffer)
+          Api.del_fvec(@out_fvec)
 
-					break
-				end
+          break
+        end
 			end
 		end
 

data/lib/aubio/pitches.rb

@@ -0,0 +1,75 @@
+module Aubio
+	class Pitches
+
+		def initialize(aubio_source, params)
+      # TODO: cleanup param dups
+			@sample_rate = params[:sample_rate] || 44100
+			@window_size = params[:window_size] || 1024
+			@hop_size    = params[:hop_size]    || 512
+
+			@source = aubio_source
+			@pitch = Api.new_aubio_pitch('yin', @window_size, @hop_size, @sample_rate)
+			Api.aubio_pitch_set_unit(@pitch, 'midi')
+			Api.aubio_pitch_set_tolerance(@pitch, 0.8)
+
+			# create output for source
+			@sample_buffer = Api.new_fvec(@hop_size)
+			# create output for pitch and beat
+			@out_fvec = Api.new_fvec(1)
+		end
+
+		def each
+			return enum_for(:each) unless block_given?
+
+			total_frames_counter = 0
+			read_buffer = FFI::MemoryPointer.new(:int)
+      last_pitch = 0
+
+			loop do
+				# Perform pitch calculation
+        Api.aubio_source_do(@source, @sample_buffer, read_buffer)
+				Api.aubio_pitch_do(@pitch, @sample_buffer, @out_fvec)
+
+        # Retrieve result
+        pitch = Api.fvec_get_sample(@out_fvec, 0)
+        confidence = Api.aubio_pitch_get_confidence(@pitch)
+        no_of_bytes_read = read_buffer.read_int
+        total_frames_counter += no_of_bytes_read
+
+        if (last_pitch - pitch).abs >= 1 and confidence > 0.9
+					output = {
+            :pitch => pitch,
+            :confidence => confidence,
+            :start => (total_frames_counter == 0 ? 1 : 0),
+            :end => 0
+					}
+          yield output
+				end
+
+        last_pitch = pitch
+
+        if no_of_bytes_read != @hop_size
+          # there's no more audio to look at
+
+          # Let's output one last pitch to mark the end of the file
+          total_time = total_frames_counter.to_f / @sample_rate.to_f
+          output = {
+            :pitch => pitch,
+            :confidence => confidence,
+            :start => 0,
+            :end  => 1
+          }
+          yield output
+
+          # clean up
+          Api.del_aubio_pitch(@pitch)
+          Api.del_fvec(@sample_buffer)
+          Api.del_fvec(@out_fvec)
+
+          break
+        end
+			end
+		end
+
+	end
+end

data/lib/aubio/version.rb

@@ -1,3 +1,3 @@
 module Aubio
-  VERSION = "0.2.0"
+  VERSION = "0.2.1"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: aubio
 version: !ruby/object:Gem::Version
-  version: 0.2.0
+  version: 0.2.1
 platform: ruby
 authors:
 - Xavier Riley
@@ -89,7 +89,9 @@ files:
 - bin/setup
 - lib/aubio.rb
 - lib/aubio/api.rb
+- lib/aubio/beats.rb
 - lib/aubio/onsets.rb
+- lib/aubio/pitches.rb
 - lib/aubio/version.rb
 homepage: https://github.com/xavriley/ruby-aubio
 licenses: