声音剂量

本文最后更新于 2024年2月1日 晚上

Android 14 通过持续监控声音剂量测量结果并向用户发出有关有害暴露水平的警告,在音频框架和音频 HAL 中为声音剂量提供支持。

声音剂量是在一段时间内的声压级测量值。通过监测声音剂量,我们可以帮助用户免受过度暴露或长时间暴露声音带来的破坏性影响,从而在便携式 Android 设备上使用耳机时提供更好的听力保护,并最大限度地减少听力受损的可能性。

安全收听设备的新标准符合 IEC62368-1 第 3 版(需要登录)和 EN50332-3(仅限订户访问)中听力保护的监管要求,其中引入了声音剂量的概念。

音效检测功能让原始设备制造商 (OEM) 能够遵守新的听力安全法规。为了支持声音剂量,OEM 必须遵循接口规范和法规,实现所有自定义和认证。自定义 OEM 实现可以绕过或修改声音音量的默认 AOSP 实现。不过,强烈建议您使用 AOSP 实现。

声音剂量计算

通过计算计算得出的声音剂量 (CSD),IEC62368-1 第 3 版和 EN50332-3 中的标准提高了测量声音暴露的准确性。计算 CSD 的方法是随着时间的推移累积短暂的曝光水平 (MEL)。为计算声音剂量,要保留连续 7 天的累积 CSD 值。

根据 IEC62368-1 第 3 版第 10.6.3.2 节,如果 CSD 值达到 100%,系统会在每次增加 100% 时提醒用户。如果用户未确认警告,音量会调低为 IEC62368-1 表 39 中预定义的辐射能量等级 1 (RS1) 值。

如 IEC62368-1 第 3 版第 10.6.3.3 节所述,每次 MEL 值超过 IEC62368 表 39 的辐射能源 2 类 (RS2) 值时,除了声音剂量警告之外,系统还必须启动基于暴露的警告 -1。

若要根据这些法规进行认证并提高 CSD 值的相关性,系统必须使用用户感知的准确输出值(例如媒体播放输出值)。对 CSD 计算而言,使用的值要与用户实际接触到的声压级值相近。

架构

根据帧的捕获位置,变频器的硬件特性和影响会影响所渲染帧的功耗级别。为获得精确的输出声压级测量结果,我们扩展了 HAL 以直接从底层硬件获取 MEL 值,并考虑到数字信号处理器 (DSP) 或扬声器属性,例如阻抗、灵敏度和频率响应。

如果 HAL 无法提供 MEL 值,则音频框架会分析和计算 CSD。音频框架中的此计算基于从 HAL 报告的渲染输出以及发送到音频 DSP 的帧的相关信息。

声音剂量引入了 SoundDoseHelperSoundDoseManager, 两个组件,如图 1 所示:

sound_dose_arch

图 1. 声音剂量功能的架构组件。

SoundDoseHelper

位于 systemserver 进程中的 SoundDoseHelper 类是所有相关声音剂量数据的主要收集点。AudioService 类用于管理 SoundDoseHelper 类。

SoundDoseHelper 类负责以下操作:

  • 处理新的剂量信息
  • 持续记录声音剂量值
  • audioserver 崩溃时恢复状态
  • 触发系统界面通知
  • 调低音量

SoundDoseManager

位于 audioserver 进程中的 SoundDoseManager 类是 AudioFlinger 类的一部分,用于从 HAL 收集声音剂量数据,或在发送到 HAL 的帧内部计算这些数据(作为后备选项)。SoundDoseManager 类会将声音剂量数据发送到 SoundDoseHelper 类。

MelProcessor 和 MelAggregator

如果 HAL 无法提供 MEL 值,则应使用 libaudioutils 中的 MelProcessorMelAggregator 实用程序用于内部声音剂量计算。

MelProcessor 类中,通过调用 MelProcessor::process(const void* buffer, size_t bytes),在具有音频样本的缓冲区上执行主计算。 OEM 可以根据需要在 HAL 实现中使用 MelProcessor

MelAggregator 类从不同的音频端口接收 MEL 值,并使用 7 天滚动时间计算 CSD 值。方法 MelAggregator::aggregateAndAddNewMelRecord_l(MelRecord mel) 会执行该逻辑。系统会将结果发送到 SoundDoseManager 类,以便与 AudioService 进行通信。

实现

从 Android 14 开始,HIDL 接口扩展已废弃,因此 AIDL 音频 HAL 中定义了名为 ISoundDose 的新 HAL 接口用于检索计算的 MEL 值和发出曝光警告。不过,对于需要更多时间来集成 AIDL 音频 HAL 的实现者,我们提供了独立的声音剂量 AIDL HAL,该 API 提供了 ISoundDoseFactory 界面。将来会将其废弃。

以下代码示例展示了 HAL 如何支持声音剂量:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
/**
* This interface provides functions related to sound exposure control required for compliance to
* EN/IEC 62368-1 3rd edition. Implementing this interface is mandatory for devices for which
* compliance to this standard is mandated and implementing audio offload decoding or other direct
* playback paths where volume control happens below the audio HAL.
*/
@VintfStability
interface ISoundDose {
/**
* Max value in dBA used for momentary exposure warnings as defined by IEC62368-1
* 3rd edition. This value represents the default RS2 upper bound.
*/
const int DEFAULT_MAX_RS2 = 100;
/** Min value of the RS2 threshold in dBA as defined by IEC62368-1 3rd edition. */
const int MIN_RS2 = 80;

/**
* Sets the RS2 upper bound used for momentary exposure warnings. Default value is
* DEFAULT_MAX_RS2 as specified in IEC62368-1 3rd edition.
*
* @param rs2ValueDbA custom RS2 upper bound to use
* @throws EX_ILLEGAL_ARGUMENT if rs2ValueDbA is greater than DEFAULT_MAX_RS2 or lower
* than MIN_RS2
*/
void setOutputRs2UpperBound(float rs2ValueDbA);

/**
* Gets the RS2 upper bound used for momentary exposure warnings.
*
* @return the RS2 upper bound in dBA
*/
float getOutputRs2UpperBound();

/**
* Registers the HAL callback for sound dose computation. If sound dose is supported
* the MEL values and exposure notifications will be received through this callback
* only. The internal framework MEL computation will be disabled.
* It is not possible to unregister the callback. The HAL is responsible to provide
* the MEL values throughout its lifecycle.
*
* @param callback to use when new updates are available for sound dose
*/
void registerSoundDoseCallback(in IHalSoundDoseCallback callback);

@VintfStability
oneway interface IHalSoundDoseCallback {
/**
* Called whenever the current MEL value exceeds the set RS2 upper bound.
*
* @param currentDbA the current MEL value which exceeds the RS2 upper bound
* @param audioDevice the audio device where the MEL exposure warning was recorded
*/
void onMomentaryExposureWarning(float currentDbA, in AudioDevice audioDevice);

@VintfStability
parcelable MelRecord {
/**
* Array of continuously recorded MEL values >= MIN_RS2 (1 per second).
* First value in the array was recorded at 'timestamp'.
*/
float[] melValues;
/**
* Corresponds to the time in seconds, as reported by CLOCK_MONOTONIC, when
* the first MEL entry in melValues was recorded. The timestamp values have
* to be consistent throughout all audio ports, equal timestamp values will
* be aggregated.
*/
long timestamp;
}

/**
* Provides a MelRecord containing continuous MEL values sorted by timestamp.
* Note that all the MEL values originate from the audio device specified by audioDevice.
* In case values from multiple devices need to be reported, the caller should execute
* this callback once for every device.
*
* @param melRecord contains the MEL values used for CSD
* @param audioDevice the audio device where the MEL values were recorded
*/
void onNewMelValues(in MelRecord melRecord, in AudioDevice audioDevice);
}
}

新的 HAL 接口会实现回调,用于告知框架瞬时曝光,并在输出级别超过 RS1 时提供 MEL 值。实现这些接口后,框架使用它们进行 CSD 报告。如果没有此回调实现,系统会使用 AudioFlinger 上的回退实现来计算 CSD 值的估算值。

注意:如果采用分流压缩音频解码,无需唤醒主机应用处理器,即可将压缩的音频数据直接流式传输到 DSP,从而优化电池用量。在此类硬件分流解码的情况下,框架只能访问压缩音频数据,无法估算 MEL 值。在这种情况下,在使用 AOSP 声音剂量实现时,设备只能遵循 EN50332-3/IEC62368-1 规定,即通过 HAL 接口提供 MEL 值。

声音剂量独立 AIDL 支持

在 OEM 能够在 AIDL 音频 HAL 中集成声音剂数之前,他们可以使用独立的 AIDL API ISoundDoseFactory 来解决此问题。ISoundDoseFactory 使用 ISoundDose 接口,如以下代码示例所示:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
@VintfStability
interface ISoundDoseFactory {
/**
* Retrieve the sound dose interface for a given audio HAL module name.
*
* If a device must comply to IEC62368-1 3rd edition audio safety requirements and is
* implementing audio offload decoding or other direct playback paths where volume control
* happens below the audio HAL, it must return an instance of the ISoundDose interface.
* The same instance must be returned during the lifetime of the HAL module.
* If the HAL module does not support sound dose, null must be returned, without throwing
* any errors.
*
* @param module for which we trigger sound dose updates.
* @return An instance of the ISoundDose interface implementation.
* @throws EX_ILLEGAL_STATE If there was an error creating an instance.
*/
@nullable ISoundDose getSoundDose(in @utf8InCpp String module);
}

声音剂量 AIDL 音频 HAL 支持

长期以来,在 AIDL 音频 HAL 中,可通过扩展 IModule 接口支持声音剂量接口,如以下代码示例所示:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
@VintfStability
interface IModule {

/**
* Retrieve the sound dose interface.
*
* If a device must comply to IEC62368-1 3rd edition audio safety requirements and is
* implementing audio offload decoding or other direct playback paths where volume control
* happens below the audio HAL, it must return an instance of the ISoundDose interface.
* The same instance must be returned during the lifetime of the HAL module.
* If the HAL module does not support sound dose, null must be returned, without throwing
* any errors.
*
* @return An instance of the ISoundDose interface implementation.
* @throws EX_ILLEGAL_STATE If there was an error creating an instance.
*/
@nullable ISoundDose getSoundDose();
}

此功能是 IEC62368-1 第 3 版和 EN50332-3 中所述的新法规的实现方式,因此没有任何面向外部的 API。

OEM 可以通过以下方式实现设备认证:实现新的 HAL 接口,为音频框架提供准确的 CSD 数据(推荐),或者通过提供自定义的声音剂量实现来实现。

启用声音剂量计算

默认情况下,AOSP 支持助听安全逻辑,以确保符合现有的 EN50332-2 和 IEC62368-1 10.6.5 标准。

在 Android 14 中,系统会默认停用声音剂量计算。

注意:请勿在 Android 14 版本中启用声音剂量计算。用于支持计算声音剂量的逻辑仅从 Android 14-QPR1 开始才实现。这种用于启用 CSD 的逻辑在 Android 14 上无法完全正常运行,并且可能会导致蓝牙设备的 CSD 不准确。

从 Android 14-QPR1 开始,请按照以下准则计算声音剂量。

  • 如果您所在的国家/地区强制执行声音剂量法规,请检查 config.xml 中的 config_safe_media_volume_enabled 是否已设置为 true

    注意:在未实施这些法规的国家/地区,OEM 可以通过 SysUI 启用或停用 CSD,例如在设置页面中添加一个切换开关或添加一项应用设置。

  • 为了符合 EN50332-3 和 IEC62368-1 10.6.3 要求,供应商必须将 config.xml 中的 config_safe_sound_dosage_enabled 标志叠加到 true。对于支持分流解码且未实现声音剂量 HAL 接口的设备,不得将 config_safe_sound_dosage_enabled 设置为 true。在这种情况下,将 config_safe_sound_dosage_enabled 设置为 true 可能会导致 CSD 值不准确以及不符合安全听力标准的认证问题。

    注意:OEM 可以通过替换 frameworks/base/core/res/res/values-mcc${XX}/config.xml 中的标志来为特定国家/地区设置 config_safe_sound_dose_enabled 标志,其中 XX 表示相应国家/地区的移动设备国家/地区代码 (MCC)。

下面的决策图描述了根据国家/地区限制和标志值确定计算 CSD 还是旧版听力安全等级(在 Android 14 之前实现)的逻辑。

enable_csd

图 2. 启用声音剂量计算(在 Android 14-QPR1 中添加了该逻辑)。

验证

在实现声音剂量的 HAL 接口时,OEM 必须针对 IModule AIDL 音频 HAL 实现的 VtsHalAudioCoreTargetTest 定义的 VTS 测试用例进行验证,或者针对独立声音剂量 AIDL HAL 实现的 VtsHalSoundDoseFactoryTargetTest 定义的 VTS 测试用例进行验证。


声音剂量
https://www.shangyexin.com/2024/09/08/audio-sound-dose/
作者
Yasin
发布于
2024年9月8日
许可协议