Add AVAudioConverter to try a new way of read data from the AVAudioEngin in live decoding. The idea is read the data in a more appealing format for iOS (float 32, 16000 Hz) and convert it (with AVAudioConverter) to the Sphinx format (int 16, 16000Hz). AVAudioConverter is only available from iOS 9.0 so the deployment traget needs to change.

Author: Bruno Berisso

TLSphinx.xcodeproj/project.pbxproj

@@ -308,7 +308,7 @@
 				GCC_WARN_UNINITIALIZED_AUTOS = YES_AGGRESSIVE;
 				GCC_WARN_UNUSED_FUNCTION = YES;
 				GCC_WARN_UNUSED_VARIABLE = YES;
-				IPHONEOS_DEPLOYMENT_TARGET = 8.3;
+				IPHONEOS_DEPLOYMENT_TARGET = 9.0;
 				MTL_ENABLE_DEBUG_INFO = YES;
 				ONLY_ACTIVE_ARCH = YES;
 				SDKROOT = iphoneos;
@@ -352,7 +352,7 @@
 				GCC_WARN_UNINITIALIZED_AUTOS = YES_AGGRESSIVE;
 				GCC_WARN_UNUSED_FUNCTION = YES;
 				GCC_WARN_UNUSED_VARIABLE = YES;
-				IPHONEOS_DEPLOYMENT_TARGET = 8.3;
+				IPHONEOS_DEPLOYMENT_TARGET = 9.0;
 				MTL_ENABLE_DEBUG_INFO = NO;
 				SDKROOT = iphoneos;
 				SWIFT_OPTIMIZATION_LEVEL = "-Owholemodule";
@@ -382,7 +382,7 @@
 				);
 				INFOPLIST_FILE = TLSphinx/Info.plist;
 				INSTALL_PATH = "$(LOCAL_LIBRARY_DIR)/Frameworks";
-				IPHONEOS_DEPLOYMENT_TARGET = 8.2;
+				IPHONEOS_DEPLOYMENT_TARGET = 9.0;
 				LD_RUNPATH_SEARCH_PATHS = "$(inherited) @executable_path/Frameworks @loader_path/Frameworks";
 				LIBRARY_SEARCH_PATHS = (
 					"$(PROJECT_DIR)/Sphinx/lib/pocketsphinx",
@@ -416,7 +416,7 @@
 				);
 				INFOPLIST_FILE = TLSphinx/Info.plist;
 				INSTALL_PATH = "$(LOCAL_LIBRARY_DIR)/Frameworks";
-				IPHONEOS_DEPLOYMENT_TARGET = 8.2;
+				IPHONEOS_DEPLOYMENT_TARGET = 9.0;
 				LD_RUNPATH_SEARCH_PATHS = "$(inherited) @executable_path/Frameworks @loader_path/Frameworks";
 				LIBRARY_SEARCH_PATHS = (
 					"$(PROJECT_DIR)/Sphinx/lib/pocketsphinx",

TLSphinx/Decoder.swift

@@ -181,14 +181,36 @@ public final class Decoder {
             throw DecodeErrors.NoAudioInputAvailable
         }
 
-        let formatIn = AVAudioFormat(commonFormat: .pcmFormatInt16, sampleRate: 44100, channels: 1, interleaved: false)
-        engine.connect(input, to: engine.outputNode, format: formatIn)
+        let mixer = AVAudioMixerNode()
+        engine.attach(mixer)
+        engine.connect(input, to: mixer, format: input.outputFormat(forBus: 0))
 
-        input.installTap(onBus: 0, bufferSize: 4096, format: formatIn, block: { (buffer: AVAudioPCMBuffer!, time: AVAudioTime!) -> Void in
+        let formatIn = AVAudioFormat(commonFormat: .pcmFormatFloat32, sampleRate: 16000, channels: 1, interleaved: false)
+        let formatOut = AVAudioFormat(commonFormat: .pcmFormatInt16, sampleRate: 16000, channels: 1, interleaved: false)
+        let bufferMapper = AVAudioConverter(from: formatIn, to: formatOut)
 
-            let audioData = buffer.toDate()
+        mixer.installTap(onBus: 0, bufferSize: 2048, format: formatIn, block: {
+            [unowned self] (buffer: AVAudioPCMBuffer!, time: AVAudioTime!) in
+
+            let sphinxBuffer = AVAudioPCMBuffer(pcmFormat: formatOut, frameCapacity: buffer.frameCapacity)
+
+            //This is needed because the 'frameLenght' default value is 0 (since iOS 10) and cause the 'convert' call
+            //to faile with an error (Error Domain=NSOSStatusErrorDomain Code=-50 "(null)")
+            //More here: http://stackoverflow.com/questions/39714244/avaudioconverter-is-broken-in-ios-10
+            sphinxBuffer.frameLength = sphinxBuffer.frameCapacity
+
+            do {
+                try bufferMapper.convert(to: sphinxBuffer, from: buffer)
+            } catch(let error as NSError) {
+                print(error)
+                return
+            }
+
+            let audioData = sphinxBuffer.toData()
             self.process_raw(audioData)
 
+            print("Process: \(buffer.frameLength) frames - \(audioData.count) bytes - sample time: \(time.sampleTime)")
+
             if self.speechState == .utterance {
 
                 self.end_utt()
@@ -202,9 +224,6 @@ public final class Decoder {
             }
         })
 
-        engine.mainMixerNode.outputVolume = 0.0
-        engine.prepare()
-
         start_utt()
 
         do {
@@ -218,8 +237,6 @@ public final class Decoder {
 
     public func stopDecodingSpeech () {
         engine.stop()
-        engine.mainMixerNode.removeTap(onBus: 0)
-        engine.reset()
         engine = nil
     }
 }