Expose custom configuration of internal AEC3

Author: skywhale

src/config.rs

@@ -101,6 +101,28 @@ impl From<HighPassFilter> for ffi::AudioProcessing_Config_HighPassFilter {
     }
 }
 
+/// [Highly Experimental] Configurations of internal AEC3 implementation.
+/// TODO(skywhale): Add more parameters from:
+/// https://gitlab.freedesktop.org/pulseaudio/webrtc-audio-processing/-/blob/master/webrtc/api/audio/echo_canceller3_config.h
+#[derive(Debug, Clone, PartialEq)]
+#[cfg_attr(feature = "derive_serde", derive(Serialize, Deserialize))]
+pub struct EchoCanceller3Config {
+    /// Number of linear filters to apply.
+    pub num_filters: usize,
+}
+
+impl Default for EchoCanceller3Config {
+    fn default() -> Self {
+        Self { num_filters: 5 }
+    }
+}
+
+impl From<EchoCanceller3Config> for ffi::EchoCanceller3ConfigOverride {
+    fn from(other: EchoCanceller3Config) -> Self {
+        Self { num_filters: other.num_filters as i32 }
+    }
+}
+
 /// AEC (acoustic echo cancellation) configuration.
 #[derive(Debug, Clone, PartialEq)]
 #[cfg_attr(feature = "derive_serde", derive(Serialize, Deserialize))]

src/lib.rs

@@ -8,7 +8,7 @@
 mod config;
 mod stats;
 
-use std::{error, fmt, sync::Arc};
+use std::{error, fmt, ptr::null, sync::Arc};
 use webrtc_audio_processing_sys as ffi;
 
 pub use config::*;
@@ -49,7 +49,15 @@ impl Processor {
     /// instantiation, however new configs can be be passed to `set_config()`
     /// at any time during processing.
     pub fn new(config: &InitializationConfig) -> Result<Self, Error> {
-        let inner = Arc::new(AudioProcessing::new(config)?);
+        Self::with_aec3_config(config, None)
+    }
+
+    /// Creates a new `Processor` with custom AEC3 configuration.
+    pub fn with_aec3_config(
+        config: &InitializationConfig,
+        aec3_config: Option<EchoCanceller3Config>,
+    ) -> Result<Self, Error> {
+        let inner = Arc::new(AudioProcessing::new(config, aec3_config)?);
         let num_samples = inner.num_samples_per_frame();
         Ok(Self {
             inner,
@@ -64,6 +72,13 @@ impl Processor {
         })
     }
 
+    /// Initializes internal states, while retaining all user settings. This should be called before
+    /// beginning to process a new audio stream. However, it is not necessary to call before processing
+    /// the first stream after creation.
+    pub fn initialize(&mut self) {
+        self.inner.initialize()
+    }
+
     /// Processes and modifies the audio frame from a capture device by applying
     /// signal processing as specified in the config. `frame` should hold an
     /// interleaved f32 audio frame, with NUM_SAMPLES_PER_FRAME samples.
@@ -182,13 +197,23 @@ struct AudioProcessing {
 }
 
 impl AudioProcessing {
-    fn new(config: &InitializationConfig) -> Result<Self, Error> {
+    fn new(
+        config: &InitializationConfig,
+        aec3_config: Option<EchoCanceller3Config>,
+    ) -> Result<Self, Error> {
+        let aec3_config = if let Some(aec3_config) = aec3_config {
+            &aec3_config.into() as *const ffi::EchoCanceller3ConfigOverride
+        } else {
+            null()
+        };
+
         let mut code = 0;
         let inner = unsafe {
             ffi::audio_processing_create(
                 config.num_capture_channels as i32,
                 config.num_render_channels as i32,
                 config.sample_rate_hz as i32,
+                aec3_config,
                 &mut code,
             )
         };
@@ -199,6 +224,10 @@ impl AudioProcessing {
         }
     }
 
+    fn initialize(&self) {
+        unsafe { ffi::initialize(self.inner) }
+    }
+
     fn process_capture_frame(&self, frame: &mut Vec<Vec<f32>>) -> Result<(), Error> {
         let mut frame_ptr = frame.iter_mut().map(|v| v.as_mut_ptr()).collect::<Vec<*mut f32>>();
         unsafe {
@@ -412,8 +441,8 @@ mod tests {
             ..InitializationConfig::default()
         };
         let mut ap = Processor::new(&config).unwrap();
-        
-        // tweak params outside of config 
+
+        // tweak params outside of config
         ap.set_output_will_be_muted(true);
         ap.set_stream_key_pressed(true);
 
@@ -427,5 +456,4 @@ mod tests {
 
         // it shouldn't crash
     }
-
 }

webrtc-audio-processing-sys/build.rs

@@ -7,6 +7,10 @@ fn out_dir() -> PathBuf {
     std::env::var("OUT_DIR").expect("OUT_DIR environment var not set.").into()
 }
 
+fn src_dir() -> PathBuf {
+    std::env::var("CARGO_MANIFEST_DIR").expect("CARGO_MANIFEST_DIR environment var not set.").into()
+}
+
 #[cfg(not(feature = "bundled"))]
 mod webrtc {
     use super::*;
@@ -62,6 +66,7 @@ mod webrtc {
         let mut include_paths = vec![
             out_dir().join("include"),
             out_dir().join("include").join("webrtc-audio-processing-1"),
+            src_dir().join("webrtc-audio-processing").join("webrtc"),
         ];
         let mut lib_paths = vec![out_dir().join("lib")];
 

webrtc-audio-processing-sys/src/wrapper.cpp

@@ -1,5 +1,12 @@
 #include "wrapper.hpp"
 
+// These definitions shouldn't affect the implementation of AEC3.
+// We are defining them to work around some build-time assertions
+// when including the internal header file of echo_canceller3.h
+#define WEBRTC_APM_DEBUG_DUMP 0
+#define WEBRTC_POSIX
+#include <modules/audio_processing/aec3/echo_canceller3.h>
+
 #include <algorithm>
 #include <memory>
 
@@ -29,6 +36,32 @@ OptionalBool from_absl_optional(const absl::optional<bool>& optional) {
   return rv;
 }
 
+webrtc::EchoCanceller3Config build_aec3_config(const EchoCanceller3ConfigOverride& override) {
+    webrtc::EchoCanceller3Config config;
+    config.delay.num_filters = override.num_filters;
+    return config;
+}
+
+class EchoCanceller3Factory : public webrtc::EchoControlFactory {
+ public:
+  explicit EchoCanceller3Factory(const webrtc::EchoCanceller3Config& config)
+      : config_(config) {}
+
+  std::unique_ptr<webrtc::EchoControl> Create(
+      int sample_rate_hz,
+      int num_render_channels,
+      int num_capture_channels) override {
+    return std::unique_ptr<webrtc::EchoControl>(new webrtc::EchoCanceller3(
+        config_,
+        sample_rate_hz,
+        num_render_channels,
+        num_capture_channels));
+  }
+
+ private:
+  webrtc::EchoCanceller3Config config_;
+};
+
 }  // namespace
 
 struct AudioProcessing {
@@ -43,9 +76,16 @@ AudioProcessing* audio_processing_create(
     int num_capture_channels,
     int num_render_channels,
     int sample_rate_hz,
+    const EchoCanceller3ConfigOverride* aec3_config_override,
     int* error) {
   AudioProcessing* ap = new AudioProcessing;
-  ap->processor.reset(webrtc::AudioProcessingBuilder().Create());
+
+  webrtc::AudioProcessingBuilder builder;
+  if (aec3_config_override != nullptr) {
+      auto* factory = new EchoCanceller3Factory(build_aec3_config(*aec3_config_override));
+      builder.SetEchoControlFactory(std::unique_ptr<webrtc::EchoControlFactory>(factory));
+  }
+  ap->processor.reset(builder.Create());
 
   const bool has_keyboard = false;
   ap->capture_stream_config = webrtc::StreamConfig(

webrtc-audio-processing-sys/src/wrapper.hpp

@@ -41,11 +41,18 @@ struct Stats {
   OptionalInt delay_ms;
 };
 
+// A slimmed-down version of webrtc::EchoCanceller3Config.
+// We can not just expose the webrtc variant as the binding loses all the default values.
+struct EchoCanceller3ConfigOverride {
+    int num_filters;
+};
+
 // Creates a new instance of AudioProcessing.
 AudioProcessing* audio_processing_create(
     int num_capture_channels,
     int num_render_channels,
     int sample_rate_hz,
+    const EchoCanceller3ConfigOverride* aec3_config_override,
     int* error);
 
 // Processes and modifies the audio frame from a capture device.