chromium/components/speech/endpointer/endpointer.h

// Copyright 2024 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

#ifndef COMPONENTS_SPEECH_ENDPOINTER_ENDPOINTER_H_
#define COMPONENTS_SPEECH_ENDPOINTER_ENDPOINTER_H_

#include <stdint.h>

#include "components/speech/endpointer/energy_endpointer.h"

class EpStatus;
class AudioChunk;

namespace speech {

// A simple interface to the underlying energy-endpointer implementation, this
// class lets callers provide audio as being recorded and let them poll to find
// when the user has stopped speaking.
//
// There are two events that may trigger the end of speech:
//
// speechInputPossiblyComplete event:
//
// Signals that silence/noise has  been detected for a *short* amount of
// time after some speech has been detected. It can be used for low latency
// UI feedback. To disable it, set it to a large amount.
//
// speechInputComplete event:
//
// This event is intended to signal end of input and to stop recording.
// The amount of time to wait after speech is set by
// speech_input_complete_silence_length_ and optionally two other
// parameters (see below).
// This time can be held constant, or can change as more speech is detected.
// In the latter case, the time changes after a set amount of time from the
// *beginning* of speech.  This is motivated by the expectation that there
// will be two distinct types of inputs: short search queries and longer
// dictation style input.
//
// Three parameters are used to define the piecewise constant timeout function.
// The timeout length is speech_input_complete_silence_length until
// long_speech_length, when it changes to
// long_speech_input_complete_silence_length.
class Endpointer { … };

}  // namespace speech

#endif  // COMPONENTS_SPEECH_ENDPOINTER_ENDPOINTER_H_