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/memory/weak_ptr.h"
+#include "base/process/process_handle.h"
+#include "base/sync_socket.h"
+#include "base/synchronization/lock.h"
+#include "base/threading/thread_checker.h"
+#include "content/common/content_export.h"
+#include "media/audio/audio_input_controller.h"
+
+namespace base {
+class CancelableSyncSocket;
+class TimeDelta;
+}  // namespace base
+
+namespace media {
+class AudioBus;
+class AudioParameters;
+}  // namespace media
+
+namespace content {
+// SpeechRecognitionAudioReceiver prepares the SharedMemory and SyncSocket.
+// Call to |Listen()| starts receiving data on an |task_runner_|'s thread.
+// Each time new data is ready to be consumed, |event_handler_->OnData()|
+// is executed on the |task_runner_|'s thread.
+class CONTENT_EXPORT SpeechRecognitionAudioReceiver {
+ public:
+  virtual ~SpeechRecognitionAudioReceiver();
+
+  // Factory method for creating and initializing the receiver.
+  // The |event_handler| and |task_runner| must outlive the receiver object.
+  // Only |foreign_memory_handle| and |foreign_socket_descriptor| are
+  // output parameters and are unsafe to use if return value is NULL.
+  // Returns the new successfully initialized object or NULL on failure.
+  static scoped_ptr<SpeechRecognitionAudioReceiver> Create(
+      const media::AudioParameters& audio_params,
+      const base::ProcessHandle peer_process_handle,
+      media::AudioInputController::EventHandler* event_handler,
+      base::SingleThreadTaskRunner* task_runner,
+      base::SharedMemoryHandle* foreign_memory_handle,
+      base::SyncSocket::TransitDescriptor* foreign_socket_descriptor);
+
+  // Starts listening by posting the first |Receive()| task to |task_runner_|;
+  void Listen();
+
+  // Stops posting new tasks to |task_runner_|.
+  void Stop();
+
+ private:
+  SpeechRecognitionAudioReceiver(
+      const media::AudioParameters& audio_params,
+      media::AudioInputController::EventHandler* event_handler,
+      base::SingleThreadTaskRunner* task_runner);
+
+  // Prepares the shared memory and the sockets. Returns true on success
+  // and false on any failure.
+  bool Initialize(
+      const base::ProcessHandle peer_process_handle,
+      base::SharedMemoryHandle* foreign_memory_handle,
+      base::SyncSocket::TransitDescriptor* foreign_socket_descriptor);
+
+  // Runs on the |task_runner_| when receiving audio from renderer.
+  void Receive(const base::TimeDelta& receive_timeout);
+
+  // The audio buffer in shared memory.
+  media::AudioInputBuffer* GetAudioBuffer();
+
+  // Accessed under |listening_lock_| to change the listening state.
+  void SetListening(const bool listening);
+
+  // Accessed under |listening_lock_| to check the listening state.
+  bool IsListening();
+
+  // Audio parameters for the incoming audio stream. Used to calculate the
+  // appropriate receive timeout and reposting the |Receive()| task.
+  // Accessed on the BrowserThread::IO and |task_runner_|'s thread.
+  const media::AudioParameters audio_params_;
+
+  // The duration of the audio chunk we get on each Receive().
+  const base::TimeDelta audio_chunk_duration_;
+
+  // The task runner used for receiving on the |receiving_socket_| and
+  // triggering OnData() and OnError() callbacks for the |event_handler_|.
+  base::SingleThreadTaskRunner* task_runner_;
+
+  // Client to be notified with OnData() and OnError() events.
+  media::AudioInputController::EventHandler* event_handler_;
+
+  // Used for handling consecutive timeouts or closed |receiving_socket_|.
+  // Accessed on the |task_runner_|'s thread.
+  uint32 dropped_packets_;
+
+  // Socket for receiving the next buffer index of audio data.
+  // Prepared on the BrowserThread and accessed on the |task_runner_|'s thread.
+  scoped_ptr<base::CancelableSyncSocket> receiving_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 to check if |Receive()| is executed on the |task_runner_|'s thread.
+  base::ThreadChecker thread_checker_;
+
+  // Utilized when posting tasks to the |task_runner_|.
+  base::WeakPtrFactory<SpeechRecognitionAudioReceiver> weak_ptr_factory_;
+
+  // Synchronization lock for |listening_|;
+  base::Lock listening_lock_;
+
+  // Used for synchronizing the BrowserThread::IO and |task_runner_|'s thread.
+  // We stop posting new tasks to the |task_runner_| when no longer listening.
+  // Accessed under |listening_lock_|.
+  bool listening_;
+
+  DISALLOW_COPY_AND_ASSIGN(SpeechRecognitionAudioReceiver);
+};
+
+}  // namespace content
+
+#endif  // CONTENT_BROWSER_SPEECH_SPEECH_RECOGNITION_AUDIO_RECEIVER_H_