Enable capture of RX features from RADE decoder (#776)

* Initial implementation of RX feature capture.

* Allow path to feature file to be specified at the command line.

* Fix compilation error after upgrading MacPorts.

* Add command line option for TX feature capture.

* Add -txfile command line argument to feed in WAV file through TX pipeline.

* Adjust scaling to match PR example.

* Opt for improved resampling audio quality.

* We don't actually need to add additional attenuation anymore.

* Switch over to soxr for further experimentation.

* Forgot change to have Windows build work.

* Update Linux build instructions.

* Fix additional compiler error.

* Update paCallbackData.h

* Remove missed code that's no longer needed.

* Update main.cpp

* Try to reduce latency.

* Another experiment to decrease latency.

* Go back to default settings.

* Fix failing ctests.

* Fix Windows packaging failures.

* Disable ctests for soxr.

* Enable SIMD for aarch64.

* Smooth out gaps in audio caused by how soxr works.

* Fix build errors.

* Ensure we're flushing out our output FIFO if we stop receiving input.

* ctests should now be fixed.

* Revert all samplerate changes. These will go in another PR.

cmake/Buildportaudio-2.0.cmake

@@ -17,7 +17,7 @@ if(NOT portaudio_POPULATED)
   list(APPEND FREEDV_PACKAGE_SEARCH_PATHS ${portaudio_BINARY_DIR})
 endif()
 
-list(APPEND FREEDV_LINK_LIBS PortAudio)
-list(APPEND FREEDV_STATIC_DEPS PortAudio)
+list(APPEND FREEDV_LINK_LIBS portaudio)
+list(APPEND FREEDV_STATIC_DEPS portaudio)
 
 include_directories(${portaudio_SOURCE_DIR}/include)

src/main.cpp

@@ -191,6 +191,10 @@ wxConfigBase *pConfig = NULL;
 // Unit test management
 wxString testName;
 wxString utFreeDVMode;
+wxString utTxFile;
+wxString utRxFile;
+wxString utTxFeatureFile;
+wxString utRxFeatureFile;
 
 // WxWidgets - initialize the application
 
@@ -306,8 +310,26 @@ void MainApp::UnitTest_()
             delete txEvent;
         });
 
-        // Transmit for 60 seconds
-        std::this_thread::sleep_for(60s);
+        if (utTxFile != "")
+        {
+            // Transmit until file has finished playing
+            SF_INFO     sfInfo;
+            sfInfo.format = 0;
+            g_sfPlayFile = sf_open((const char*)utTxFile.ToUTF8(), SFM_READ, &sfInfo);
+            g_sfTxFs = sfInfo.samplerate;
+            g_loopPlayFileToMicIn = false;
+            g_playFileToMicIn = true;
+
+            while (g_playFileToMicIn)
+            {
+                std::this_thread::sleep_for(20ms);
+            } 
+        }
+        else
+        {
+            // Transmit for 60 seconds
+            std::this_thread::sleep_for(60s);
+        }
 
         // Stop transmitting
         log_info("Firing PTT");
@@ -323,22 +345,40 @@ void MainApp::UnitTest_()
         sim.MouseClick();*/
 
         // Wait 5 seconds for FreeDV to stop
-        std::this_thread::sleep_for(5s);
+        //std::this_thread::sleep_for(5s);
     }
     else
     {
-        // Receive for 60 seconds
-        auto sync = 0;
-        for (int i = 0; i < 60*10; i++)
+        if (utRxFile != "")
         {
-            std::this_thread::sleep_for(100ms);
-            auto newSync = freedvInterface.getSync();
-            if (newSync != sync)
+            // Receive until file has finished playing
+            SF_INFO     sfInfo;
+            sfInfo.format = 0;
+            g_sfPlayFileFromRadio = sf_open((const char*)utRxFile.ToUTF8(), SFM_READ, &sfInfo);
+            g_sfFs = sfInfo.samplerate;
+            g_loopPlayFileFromRadio = false;
+            g_playFileFromRadio = true;
+
+            while (g_playFileFromRadio)
             {
-                log_info("Sync changed from %d to %d", sync, newSync);
-                sync = newSync;
+                std::this_thread::sleep_for(20ms);
             }
-        } 
+        }
+        else
+        {
+            // Receive for 60 seconds
+            auto sync = 0;
+            for (int i = 0; i < 60*10; i++)
+            {
+                std::this_thread::sleep_for(100ms);
+                auto newSync = freedvInterface.getSync();
+                if (newSync != sync)
+                {
+                    log_info("Sync changed from %d to %d", sync, newSync);
+                    sync = newSync;
+                }
+            } 
+        }
     }
 
     // Fire event to stop FreeDV
@@ -369,6 +409,10 @@ void MainApp::OnInitCmdLine(wxCmdLineParser& parser)
     parser.AddOption("f", "config", "Use different configuration file instead of the default.");
     parser.AddOption("ut", "unit_test", "Execute FreeDV in unit test mode.");
     parser.AddOption("utmode", wxEmptyString, "Switch FreeDV to the given mode before UT execution.");
+    parser.AddOption("rxfile", wxEmptyString, "In UT mode, pipes given WAV file through receive pipeline.");
+    parser.AddOption("txfile", wxEmptyString, "In UT mode, pipes given WAV file through transmit pipeline.");
+    parser.AddOption("rxfeaturefile", wxEmptyString, "Capture RX features from RADE decoder into the provided file.");
+    parser.AddOption("txfeaturefile", wxEmptyString, "Capture TX features from FARGAN encoder into the provided file.");
 }
 
 bool MainApp::OnCmdLineParsed(wxCmdLineParser& parser)
@@ -403,6 +447,26 @@ bool MainApp::OnCmdLineParsed(wxCmdLineParser& parser)
         {
             log_info("Using mode %s for tests", (const char*)utFreeDVMode.ToUTF8());
         }
+
+        if (parser.Found("rxfile", &utRxFile))
+        {
+            log_info("Piping %s through RX pipeline", (const char*)utRxFile.ToUTF8());
+        }
+
+        if (parser.Found("txfile", &utTxFile))
+        {
+            log_info("Piping %s through TX pipeline", (const char*)utTxFile.ToUTF8());
+        }
+    }
+
+    if (parser.Found("rxfeaturefile", &utRxFeatureFile))
+    {
+        log_info("Capturing RADE RX features into file %s", (const char*)utRxFeatureFile.ToUTF8());
+    }
+
+    if (parser.Found("txfeaturefile", &utTxFeatureFile))
+    {
+        log_info("Capturing RADE TX features into file %s", (const char*)utTxFeatureFile.ToUTF8());
     }
 
     return true;

src/pipeline/RADEReceiveStep.cpp

@@ -25,11 +25,14 @@
 #include "../defines.h"
 #include "lpcnet.h" // from Opus source tree
 
+extern wxString utRxFeatureFile;
+
 RADEReceiveStep::RADEReceiveStep(struct rade* dv, FARGANState* fargan)
     : dv_(dv)
     , fargan_(fargan)
     , inputSampleFifo_(nullptr)
     , outputSampleFifo_(nullptr)
+    , featuresFile_(nullptr)
 {
     // Set FIFO to be 2x the number of samples per run so we don't lose anything.
     inputSampleFifo_ = codec2_fifo_create(rade_nin_max(dv_) * 2);
@@ -38,10 +41,21 @@ RADEReceiveStep::RADEReceiveStep(struct rade* dv, FARGANState* fargan)
     // Enough for one second of audio. Probably way overkill.
     outputSampleFifo_ = codec2_fifo_create(16000);
     assert(outputSampleFifo_ != nullptr);
+
+    if (utRxFeatureFile != "")
+    {
+        featuresFile_ = fopen((const char*)utRxFeatureFile.ToUTF8(), "wb");
+        assert(featuresFile_ != nullptr);
+    }
 }
 
 RADEReceiveStep::~RADEReceiveStep()
 {
+    if (featuresFile_ != nullptr)
+    {
+        fclose(featuresFile_);
+    }
+
     if (inputSampleFifo_ != nullptr)
     {
         codec2_fifo_free(inputSampleFifo_);
@@ -93,6 +107,11 @@ std::shared_ptr<short> RADEReceiveStep::execute(std::shared_ptr<short> inputSamp
 
             // RADE processing (input signal->features).
             nout = rade_rx(dv_, features_out, input_buf_cplx);
+            if (featuresFile_)
+            {
+                fwrite(features_out, sizeof(float), nout, featuresFile_);
+            }
+
             for (int i = 0; i < nout; i++)
             {
                 pendingFeatures_.push_back(features_out[i]);

src/pipeline/RADEReceiveStep.h

@@ -23,6 +23,7 @@
 #ifndef AUDIO_PIPELINE__RADE_RECEIVE_STEP_H
 #define AUDIO_PIPELINE__RADE_RECEIVE_STEP_H
 
+#include <cstdio>
 #include <vector>
 #include "IPipelineStep.h"
 #include "../freedv_interface.h"
@@ -53,6 +54,8 @@ class RADEReceiveStep : public IPipelineStep
     struct FIFO* inputSampleFifo_;
     struct FIFO* outputSampleFifo_;
     std::vector<float> pendingFeatures_;
+
+    FILE* featuresFile_;
 };
 
 #endif // AUDIO_PIPELINE__RADE_RECEIVE_STEP_H

src/pipeline/RADETransmitStep.cpp

@@ -23,23 +23,38 @@
 #include <cstring>
 #include <cassert>
 #include <cmath>
+#include "../defines.h"
 #include "codec2_fifo.h"
 #include "RADETransmitStep.h"
 
+extern wxString utTxFeatureFile;
+
 RADETransmitStep::RADETransmitStep(struct rade* dv, LPCNetEncState* encState)
     : dv_(dv)
     , encState_(encState)
     , inputSampleFifo_(nullptr)
     , outputSampleFifo_(nullptr)
+    , featuresFile_(nullptr)
 {
     inputSampleFifo_ = codec2_fifo_create(RADE_SPEECH_SAMPLE_RATE);
     assert(inputSampleFifo_ != nullptr);
     outputSampleFifo_ = codec2_fifo_create(RADE_MODEM_SAMPLE_RATE);
     assert(outputSampleFifo_ != nullptr);
+
+    if (utTxFeatureFile != "")
+    {
+        featuresFile_ = fopen((const char*)utTxFeatureFile.ToUTF8(), "wb");
+        assert(featuresFile_ != nullptr);
+    }
 }
 
 RADETransmitStep::~RADETransmitStep()
 {
+    if (featuresFile_ != nullptr)
+    {
+        fclose(featuresFile_);
+    }
+
     if (inputSampleFifo_ != nullptr)
     {
         codec2_fifo_free(inputSampleFifo_);
@@ -86,6 +101,12 @@ std::shared_ptr<short> RADETransmitStep::execute(std::shared_ptr<short> inputSam
             // Feature extraction
             codec2_fifo_read(inputSampleFifo_, pcm, LPCNET_FRAME_SIZE);
             lpcnet_compute_single_frame_features(encState_, pcm, features, arch);
+
+            if (featuresFile_)
+            {
+                fwrite(features, sizeof(float), NB_TOTAL_FEATURES, featuresFile_);
+            }
+
             for (int index = 0; index < NB_TOTAL_FEATURES; index++)
             {
                 featureList_.push_back(features[index]);
@@ -102,7 +123,7 @@ std::shared_ptr<short> RADETransmitStep::execute(std::shared_ptr<short> inputSam
                 for (int index = 0; index < numOutputSamples; index++)
                 {
                     // We only need the real component for TX.
-                    radeOutShort[index] = radeOut[index].real * 32767;
+                    radeOutShort[index] = radeOut[index].real * 16383;
                 }
                 codec2_fifo_write(outputSampleFifo_, radeOutShort, numOutputSamples);
             }

src/pipeline/RADETransmitStep.h

@@ -23,6 +23,7 @@
 #ifndef AUDIO_PIPELINE__RADE_TRANSMIT_STEP_H
 #define AUDIO_PIPELINE__RADE_TRANSMIT_STEP_H
 
+#include <cstdio>
 #include <vector>
 #include "IPipelineStep.h"
 #include "../freedv_interface.h"
@@ -48,6 +49,8 @@ class RADETransmitStep : public IPipelineStep
     struct FIFO* inputSampleFifo_;
     struct FIFO* outputSampleFifo_;
     std::vector<float> featureList_;
+
+    FILE* featuresFile_;
 };
 
 #endif // AUDIO_PIPELINE__RADE_TRANSMIT_STEP_H

src/pipeline/ResampleStep.cpp

@@ -73,7 +73,7 @@ ResampleStep::ResampleStep(int inputSampleRate, int outputSampleRate)
     , outputSampleRate_(outputSampleRate)
 {
     int src_error;
-    resampleState_ = src_new(SRC_SINC_FASTEST, 1, &src_error);
+    resampleState_ = src_new(SRC_SINC_MEDIUM_QUALITY, 1, &src_error);
     assert(resampleState_ != nullptr);
 }
 

src/pipeline/TxRxThread.cpp

@@ -268,11 +268,13 @@ void TxRxThread::initializePipeline_()
         auto txAttenuationStep = new LevelAdjustStep(outputSampleRate_, []() {
             double dbLoss = g_txLevel / 10.0;
 
+#if 0
             if (freedvInterface.getTxMode() == FREEDV_MODE_RADE)
             {
                 // Attenuate by 4 dB as there's no BPF; anything louder distorts the signal
                 dbLoss -= 4.0;
             }
+#endif // 0
 
             double scaleFactor = exp(dbLoss/20.0 * log(10.0));
             return scaleFactor;