android-sensors

name	android-sensors
description	26+安卓传感器速查手册（vivo S19检测到43个，本文档覆盖26个有实用价值的常见传感器）。使用时机：需要查传感器列表、理解传感器用途、组合传感器推断设备状态/用户活动/环境。| 26+ Android sensor reference (43 detected on vivo S19, 26 commonly useful sensors documented here). Use when: need sensor listing, understanding sensor use cases, combining sensors for device state inference.

Android Sensors — Claude Code Skill

Complete reference for accessing Android sensors via ADB/Termux (43 detected on vivo S19, 26 commonly useful sensors documented here). Sensor count varies by device — typical is 20-30. Sensors expose raw hardware data that can be combined to infer device state, user activity, and environment — without app-level APIs.

Quick Commands

# Full sensor list (43 sensors on vivo S19, varies by device)
adb shell dumpsys sensorservice

# Active sensors only (currently being read by some process)
adb shell dumpsys sensorservice | grep -E "0x[0-9a-f]{8}\)" | head -20

# Specific sensor details
adb shell dumpsys sensorservice | grep -A 5 "Accelerometer"
adb shell dumpsys sensorservice | grep -A 5 "Ambient Light"
adb shell dumpsys sensorservice | grep -A 5 "Proximity"
adb shell dumpsys sensorservice | grep -A 1 "step_counter"

Sensor Catalog

Motion Sensors

Sensor	Raw Data	Use Case
Accelerometer	3-axis (x/y/z) m/s²	Device orientation: z≈9.8 = flat/lying down. Rhythmic spikes = walking. Dense fast spikes = running. Combined with gyro to distinguish gaming vs scrolling.
Gyroscope	Rotation rate (rad/s)	Screen rotation speed. Rapid spins = gaming. Stable axis + accelerometer = watching video lying down.
Gravity	Gravity vector (x/y/z)	Cleaner orientation than accelerometer. z-axis downward = lying flat. z-axis sideways = sitting upright.
Linear Acceleration	Acceleration minus gravity	Pure body movement. Gait frequency can estimate walking vs running.

Orientation Sensors

Sensor	Raw Data	Use Case
Rotation Vector	3D orientation quaternion	Screen facing up = lying down scrolling. Screen facing down = phone flipped over (possibly ignoring notifications).
Orientation	Angular deflection	Portrait = messaging. Landscape (90°) = watching video or gaming.
Window Orientation	System-level rotation	0=portrait, 90=landscape. Device-specific (vivo), more reliable than raw orientation sensor.

Environmental Sensors

Sensor	Raw Data	Use Case
Ambient Light	Lux	Key sensor. lux < 5 = lights off in room. Sudden increase = lights turned on or daybreak. Combine with time to detect late-night screen use in the dark.
Proximity	Distance (cm)	0 = against ear (on a call). 5cm+ = nothing close. Combine with accelerometer: lying+prox=N = scrolling; lying+prox=0 = phone call.
Sensor Temperature	Chip temperature (°C)	>40°C = heavy GPU load (gaming or charging). Can estimate session intensity.

Activity Recognition (vivo-specific)

Sensor	Raw Data	Use Case
Step Counter	Total steps since boot	Daily activity baseline. >15000 = very active. <500 = sedentary. Read from health data file for persistence.
Step Detector	Trigger per step	Real-time walking detection — fires on each step, not a cumulative counter.
vivo_activity	Activity class enum	Most useful sensor. Outputs: walking/running/cycling/in_vehicle/stationary. Direct activity classification — no manual inference needed.
Elevator Detect	Boolean trigger	In an elevator → transitioning between floors. Combine with time of day to infer going to class or returning.
Car Navi Detect	Boolean trigger	In a moving vehicle. Combined with regular accelerometer patterns = traveling.

Event Detection

Sensor	Trigger Pattern	Use Case
Raise-up (Wakeup)	Pick up → screen on	Key timing signal. Device picked up and screen turned on — user is about to look at the phone. Optimal moment for notifications.
Put-down	Device set down	User put the phone down — may have switched to another task. Don't expect immediate response.
Stationary Detect	No movement timeout	Device has not moved. Long duration = phone left untouched — user may be sleeping, in class, or occupied.
Motion Detect	Movement starts	Device began moving again after being stationary. User returned to phone.
Significant Motion	Large movement event	Major repositioning — picked up from table, got out of bed. Not micro-adjustments from scrolling.

Specialty Sensors

Sensor	Raw Data	Use Case
Drop Depth	Free-fall detection	Phone dropped. May auto-trigger screen-off protection.
Motion Sickness	Motion comfort index	User experiencing motion sickness in vehicle — minimize notification onslaught.
Game Gesture	Gaming mode trigger	Active gaming session detected. Notifications may be intrusive — defer non-urgent messages.
Angle Judge	Screen tilt angle	Flat vs upright vs 45°. Combine with light: flat+dark = side-lying phone use.
Raise-up Wakeup	Pick-up auto-wake	Device setting: pick up to wake screen. Triggers without unlock — user glances at lock screen notifications.
Smart Prox	Enhanced proximity	Device-specific (vivo). Distinguishes ear/pocket/table vs generic near/far.

Composite Inference Patterns

Combine sensors for higher-confidence state detection:

Is the user actually working/studying?

vivo_activity = stationary + screen occasionally on + no gaming gestures + not walking/running
↓
Probably sitting and working, not gaming or commuting.

Late-night screen time

time ∈ [00:00, 06:00] + ambient_light < 5 lux + screen = Awake
↓
Dark room + awake = late-night phone use. Fire notification or log for wellness tracking.

On a phone call

proximity = 0 + accelerometer = relatively_stable + no gaming gestures
↓
Phone against ear, not moving much = on a call. Avoid interruptions.

Left the phone behind

stationary_detect > 30min + step_counter incrementing (from health_data)
↓
Phone hasn't moved but steps are accumulating elsewhere = user walked away without phone.

Gaming session

game_gesture = active + gyroscope = rapid_rotation + sensor_temp > 38°C
↓
Active gaming with gyro controls, device heating up. Defer non-urgent notifications.

Technical Limitations

• Real-time raw values: Android does not expose sensor buffer reads via command line. dumpsys sensorservice shows registration status and metadata, not live data streams.
• Activity recognition (vivo_activity, Elevator Detect, etc.): Vendor-specific. Availability varies by manufacturer (vivo, Xiaomi, Samsung, etc.). Test on target device.
• Step counter: Prefer health data aggregators (Gadgetbridge, Google Fit) over raw sensor — they handle dedup and persistence.
• Phone API extension: For near-real-time sensor access, implement a Shizuku-based sensor polling endpoint in your phone-api layer.

Tested Device

vivo S19 (2026-07-13): 43 sensors detected, all responsive via dumpsys sensorservice. Sensor availability and naming varies by manufacturer and Android version. Always verify with dumpsys sensorservice on the target device.