With the maturation of large model capabilities and breakthroughs in multimodal interaction technology, AI glasses are no longer just a lightweight branch of AR/VR; they are being redefined as the "next-generation core hub for human-computer interaction." User expectations for AI glasses have evolved beyond simply "hearing" to a four-level hierarchy: "From hearing, to audibility, to clarity, to comprehension":
· Hearing – The microphone completes basic sound capture.
· Audibility – Ensuring sufficient volume and undistorted original sound.
· Clarity – Effectively separating human voice from ambient noise.
· Comprehension – Relying on AI to accurately interpret user intent.
Currently, most devices only reach the second stage, with a few achieving voice noise separation. Very few products enable AI to accurately understand commands. Full-link system-level synergy is becoming the new industry standard.
AWINIC is deeply exploring user pain points and empowering leading AI glasses products in the industry. With its core technology of "uplink capture – mid-end optimization – downlink output – full-link synergy," AWINIC has connected all four stages. By deploying a rich portfolio of product categories, AWINIC is driving smart audio devices from merely "producing sound" to "sensing and enabling interaction." How does AWINIC achieve this? Through "uplink + downlink" algorithms.
I. AI Glasses Uplink Audio Solution – Dijiang™ X1
Uplink Audio: Sound Capture and Upload
01 Pinpointing the Pain Points:
User expectations for essential scenarios such as real-time translation, first-person recording, AR navigation, and accessibility assistance are continuously rising. Can you record a vlog outdoors in high winds? Can you make a phone call in a noisy environment? Can you hear commands clearly amid the rumble of the subway? AWINIC is deeply exploring user pain points and empowering leading AI glasses products in the industry.
02 The Key to Breaking Through: AWINIC Dijiang™ Uplink Algorithm
AWINIC Dijiang™ is a series of uplink audio algorithms launched by AWINIC, targeting scenarios such as recording, video capture, and calling. It offers multiple solution suites, covering core algorithms including wind noise reduction, surround sound, noise cancellation, echo cancellation, and beamforming. It supports integration with various mainstream platforms, allowing flexible combination of algorithm modules based on different scenario needs, comprehensively empowering AI glasses for diverse use cases such as outdoor travel, conference calls, and daily recording.
1. 🌬️ Vlog Scenario Empowerment: Fearless Against Motion Wind Noise, Voice Remains Clear and Transparent
Have you ever experienced such regrets?
🚴🏻♀️ Cycling, with
the wind roar drowning out your inner monologue;
🏃🏻♀️ Running while
filming, gasping and voice becoming indistinguishable;
🚶🏻♀️ Walking
outdoors, rich ambient sound but missing the clarity of that sentence, "I
want to tell you"?
Figure 1: Motion Wind Noise Scenario Demonstration
🌀 To address this, AWINIC Dijiang™ has developed a new proprietary wind noise algorithm for AI glasses:
Acoustic signals captured by the microphone array pass through the wind noise algorithm, which accurately identifies wind noise and enhances speech clarity. Then, the surround sound module enhances the sense of immersion, redefining the sound aesthetic of vlogs.
With it, even strong winds are not a problem — every frame is matched with clear, warm sound.
· Status Detection: Transmits noise flag.
· Optional Modules (within dashed circles): Non-essential, for lightweight needs.
· Implemented (AWINIC blue background): Implemented modules.
Figure 2: Vlog Scenario Algorithm Block Diagram
Effect Demonstration
① Wind Noise Algorithm Excels in Different Environments
✅ No wind
& Light wind | Intelligently builds an immersive surround sound
field, giving everyday conversations a cinematic sense of space.
✅ Strong wind | Preserves ambient sound
while improving voice signal SNR. Not "muting," but allowing the
voice to emerge from the noise, turning the atmosphere into texture.
2. 🌐 Full-Scenario Call Empowerment: Intelligent Noise Cancellation, Precise Voice Transmission
Have you ever had moments like these?
💻 During a video conference, your voice sounds muffled,
and a colleague frowns, asking, "What did you say?"
🕶 Taking an important call on a busy street, with traffic
roar and horns blaring, the other person only hears a "buzz——";
📝 Communicating abroad in a noisy environment, a clerk
enthusiastically introduces spinach: "Do you like spinach?" but the
translation tool interprets it as "You look like a Spaniard"...
Figure 3: Translation Scenario Demonstration
📞 To address this, AWINIC Dijiang™ delves deep into the full-link acoustic scene of calls:
Acoustic signals captured by the microphone array are precisely stripped of echo signals by the echo cancellation module. Beamforming acts like an invisible spotlight for sound, dynamically locking onto the sound source direction and narrowing the effective pickup area. Finally, noise cancellation blocks external noise, delivering an ultimate calling experience with minimal voice distortion.
· Status Detection: Transmits noise flag.
· Optional Modules (within dashed circles): Non-essential, for lightweight needs.
· Implemented (AWINIC blue background): Implemented modules.
Figure 4: Call Scenario Algorithm Block Diagram
3. 🗣 The "First Neural Hub" for Wake-up and Recognition
Have you ever experienced these moments?
🚇 Wearing glasses on the subway, trying to ask about the
weather, but the wind noise drowns out your voice;
☕ Chatting with a friend in a café, you just say "Hey——,"
and the AI glasses mistakenly wake up;
🕺🏻 Strolling and casually calling out, but
the system takes two seconds to respond...
*Figure 5: Voice Wake-up Scenario Demonstration*
🎤 So AWINIC Dijiang™ arrives. The front-end voice gatekeeper designed specifically for AI glasses:
It enhances the speech signal-to-noise ratio in complex environments (wind noise, human chatter, reverberation). In real-world wearing scenarios, recognition stability is significantly improved, with a character error rate reduction of over 6%.
· Optional Modules (within dashed circles): Non-essential, for lightweight needs.
· Planned (gray background): Future roadmap.
· Implemented (AWINIC blue background): Implemented modules.
*Figure 6: Wake-up Recognition Scenario Algorithm Block Diagram*
⏳ Waking up is no longer just about "being able to trigger." Not waking up is frustrating, false triggers are awkward, and slow responses are exhausting. User experience is the sole judge. In the future, AWINIC will develop voice wake-up algorithms with ultra-low power consumption and ultra-high wake-up rates — quieter, sharper. After all, the best interaction is one where you don't even realize it's working.
II. AI Glasses Downlink Audio Solution – awinicSKTune® Immortal Algorithm W1
Downlink Audio: Sound Playback and Output
01 Pinpointing the Pain Points:
Speakers in AR glasses are typically placed in the temple arms. For aesthetics and portability, the cavity space is extremely tight. The driver weighs less than 2 grams, with dimensions ≤10×18mm and thickness ≤3.5mm. Due to these physical constraints, such micro-speakers suffer from weak volume and low-frequency performance. Dual drivers produce sound independently, making it difficult to achieve a surround sound field, while also being prone to noticeable airflow artifacts. Therefore, when AI glasses play music, the sound quality is thin and weak, lacking bass and immersive surround sound. How can this be solved?
Figure 7: Speaker Placement Diagram (Single Side)
02 The Key to Breaking Through: awinicSKTune® Immortal Algorithm W1
AWINIC's awinicSKTune® Immortal Algorithm W1, with its exceptionally simple yet effective algorithmic performance, is the core key to solving these problems.
Figure 8: awinicSKTune® Immortal Algorithm W1 Sound Processing
Figure 9: Traditional Sound Processing
The awinicSKTune® Immortal Algorithm W1 helps smart wearable manufacturers achieve superior low-frequency performance, lower distortion, and a more immersive audio experience even within compact layout designs.
· AI Sound Field Surround Technology:
Uses AI element recognition to separate and control different audio components,
then renders the positions of virtual sound sources, simulating the effect of
sound arriving at your ears "from different directions and
distances."
Figure 10: AI Sound Field Surround Technology
· Bass Enhancement Technology:
Due to their small size and light weight, common AI glasses speakers have a
limited tolerance for low-frequency voltage signals.
Figure 11: Typical AI Glasses EQ Voltage Curve
Traditional processing methods only use EQ high-pass or low-shelf filters for pre-processing to reduce low-frequency energy and avoid mechanical distortion of the speaker diaphragm. This method inevitably compromises the overall low-frequency effect, especially the critical 50Hz-200Hz range.
The Bass Enhancement technology in awinicSKTune® Immortal Algorithm W1 provides a complete bass enhancement solution. It establishes a speaker displacement model curve to ensure all signals operate within a safe amplitude range. Then, it applies differentiated bass enhancement techniques, balancing the virtual component perception of large and small signals to improve the low-frequency performance of drums and vocals.
· Nonlinear Distortion Suppression Algorithm:
Due to magnetic circuit nonlinearity, suspension system nonlinearity, and
breakup modes at large amplitudes, speakers are prone to nonlinear distortion
at high amplitudes. This causes low-frequency buzzing, reduced clarity, and
negatively impacts the listening experience and low-frequency performance. The
nonlinear distortion suppression algorithm repairs low-frequency perception.
Combined with bass enhancement technology, it maintains pure timbre while
increasing low-frequency dynamics.
*Figure 12: Distortion Comparison with NEC Algorithm On/Off (Same Input)*
Figure 13: Input Level Comparison at Equivalent Distortion Levels
· Piano Noise Suppression Algorithm:
The APR technology in awinicSKTune® Immortal Algorithm W1 uses AI to
intelligently identify playback audio elements, accurately determining whether
the source will produce airflow artifacts. With flexible processing methods, it
solves the problem of speaker airflow artifacts and piano noise by providing
dynamic compression capability of over 6dB, without sacrificing other audio
elements or bass effects.
Figure 14: AWINIC AI Artifact Noise Suppression Algorithm
· Intelligent Volume Control Algorithm:
Poor audibility at high outdoor volumes and lack of low-frequency perception at
low indoor/medium volumes are also common pain points for glasses products. The
intelligent volume control algorithm in awinicSKTune® Immortal Algorithm W1
adjusts the EQ curve in real-time based on volume level information from the
platform side. At low volumes, human ear sensitivity to low frequencies
decreases, so the algorithm automatically boosts low-frequency gain. At high
volumes, to prevent speaker overload, it automatically reduces low-frequency
gain, boosts mid-frequency (voice) clarity, and automatically compresses peaks
to reduce artifacts.
One-click switching and separate tuning ensure the best sound for every scenario.
Figure 15: Examples of Tuning Styles Under Different Modes
o Blue Curve (Indoor Sound Quality Mode): Typically used for scenarios such as music listening. Features balanced performance across low, mid, and high frequencies, authentically restoring sound details and layering.
o Yellow Curve (Extra-Loud Volume Mode): Typically used for noisy outdoor scenarios. Boosts mid and high frequencies, significantly enhancing voice clarity and penetration.
o Other desired scenarios can also be defined based on specific requirements.
Furthermore, the awinicSKTune® Immortal Algorithm W1 has been successfully ported and functionally verified on major platforms, making it the preferred audio solution for wearable products.
Good sound should not be limited by size. As a leading player in both analog and digital technologies, AWINIC Electronics is committed to empowering next-generation smart wearables with acoustic algorithms, delivering a high-quality audio experience to users.
沪ICP备10011954号 |
沪公网安备 31011202011955号
