Binding media stream audio track to speech recognition [browser]

content/browser/speech/speech_recognition_audio_receiver.h

+// Copyright 2014 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef CONTENT_BROWSER_SPEECH_SPEECH_RECOGNITION_AUDIO_RECEIVER_H_
+#define CONTENT_BROWSER_SPEECH_SPEECH_RECOGNITION_AUDIO_RECEIVER_H_
+
+#include "base/basictypes.h"
+#include "base/memory/scoped_ptr.h"
+#include "base/memory/shared_memory.h"
+#include "base/process/process_handle.h"
+#include "base/sync_socket.h"
+#include "base/threading/thread.h"
+#include "content/common/content_export.h"
+#include "content/public/browser/browser_thread.h"
+
+namespace base {
+class Thread;
+class ThreadChecker;
+class TimeDelta;
+class CancelableSyncSocket;
+}  // namespace base
+
+namespace media {
+class AudioBus;
+class AudioInputController;
+class AudioParameters;
+}  // namespace media
+
+namespace content {
+
+// The SpeechRecognitionAudioReceiver prepares the SharedMemory and SyncSocket
+// and starts receiving data on an |audio_thread_|.
+// Once the |audio_socket_| receives a buffer index, it posts a task on the
+// browser thread to run the |on_data_cb_|.
+// The audio bus can be accessed on the browser thread via |audio_bus()|.
+// Client should check if the receive started with |started()|.
+class CONTENT_EXPORT SpeechRecognitionAudioReceiver {
+ public:
+  // OnDataCB matches AudioInputController::EventHandler::OnData.
+  // Although the AudioInputController is not used (passed as NULL),
+  // this is a handy signature to easily attach to an existing
+  // AudioInputController::EventHandler implementation.
+  typedef base::Callback<void(media::AudioInputController*,
+                              const media::AudioBus*)> OnDataCB;
+
+  SpeechRecognitionAudioReceiver(
+      const media::AudioParameters& params,
+      const base::ProcessHandle peer_process_handle,
+      const base::TimeDelta receive_timeout,
+      const OnDataCB& on_data_callback,
+      base::SharedMemoryHandle* foreign_memory_handle,
+      base::SyncSocket::TransitDescriptor* foreign_socket_descriptor);
+
+  virtual ~SpeechRecognitionAudioReceiver();
+
+  // Returns true if successfully initialized and posted the receiving task.
+  virtual bool listening();
+
+ private:
+  // Runs on separate |audio_thread_| when receiving audio from renderer.
+  virtual void OnData();
+
+  // Used to check if everything was initialized and audio thread started.
+  bool listening_;
+
+  // The timeout for the |audio_socket_| receive.
+  const base::TimeDelta receive_timeout_;
+
+  // Callback to run once audio data is ready. Runs on the |audio_thread_|.
+  OnDataCB on_data_cb_;
+
+  // Used for handling consecutive timeouts or detect closed |audio_socket_|.
+  // Accessed on the |audio_thread|.
+  uint32 dropped_packets_;
+
+  // Socket for receiving a buffer index.
+  // Prepared on the BrowserThread and accessed on the |audio_thread_|.
+  // No lock is required.
+  scoped_ptr<base::CancelableSyncSocket> audio_socket_;
+
+  // Socket for the peer process (renderer).
+  scoped_ptr<base::CancelableSyncSocket> foreign_socket_;
+
+  // Shared memory for passing audio data from renderer to the browser.
+  scoped_ptr<base::SharedMemory> shared_memory_;
+
+  // Audio bus wrapping the shared memory from the renderer.
+  // Accessed on BrowserThread.
+  scoped_ptr<media::AudioBus> audio_bus_;
+
+  // Used for synchronization when consuming audio on the |audio_bus_|.
+  // Initialized on the BrowserThread and accessed on the |audio_thread_|.
+  // Since no further reads/writes are done on the browser thread after
+  // initialization, no lock is required.
+  uint32* buffer_index_;
+
+  // Thread for receiving audio buffer indices from the renderer.
+  scoped_ptr<base::Thread> audio_thread_;
+
+  // Used to check if |OnData()| is executed on |audio_thread_|.
+  base::ThreadChecker audio_thread_checker_;
+
+  DISALLOW_COPY_AND_ASSIGN(SpeechRecognitionAudioReceiver);
+};
+
+}  // namespace content
+
+#endif  // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNITION_AUDIO_RECEIVER_H_