Real-World Applications
mpWAV provides the integrated voice interface stack — single-microphone noise removal, multi-microphone echo and noise preprocessing, wake-word detection, sound source localization, speech separation, speaker diarization, on-device speech recognition, and a conversational LLM.
For product environments as different as robots, kiosks, vehicles, appliances, smart devices, and meeting systems, we select the required technologies and connect them into one working architecture.
Technology Matched to the Environment
mpWAV analyzes your product's microphone count, speaker output, user position, number of speakers, noise environment, and service goal to compose the right combination of voice technologies.
Earbuds with limited microphone options start from mpNC; multi-channel robots from mpAEC, mpBeamforming, and mpLocalization; meetings from mpSeparation and mpDiarization; conversational kiosks from mpASR and mpLLM.
We don't force one generic algorithm onto every product — we design the voice stack around the product's conditions.
Application Overview
Multi-channel voice interfaces for products that move and speak
Helps robots detect the user's call, find the direction the voice came from, and accurately recognize commands amid their own speakers and motor noise.
Voice ordering and on-device dialogue in store noise
Captures the user's voice with a multi-channel array, removes store noise and prompt echo, then understands order intent and options through recognition and a conversational language model.
Voice commands amid driving noise and car audio
Removes car audio echo and road noise, then analyzes the wake word, seat or speech direction, and voice commands.
Single-microphone processing for products with limited mic options
For earbuds, wearables, and small smart devices where an array isn't possible, one microphone is enough to remove ambient noise.
Separating voices and organizing them by speaker
Separates multi-party and overlapping speech by speaker, tells who spoke when, then connects meetings to transcripts and summaries.
Detecting abnormal sounds amid complex equipment noise
Analyzes acoustic signals so abnormal motor and equipment sounds can be detected even on noisy production lines.
Voice interfaces for extremely noisy environments
Implements voice preprocessing, wake-word detection, on-device recognition, and dedicated hardware for extreme ambient noise and constrained networks.
Smoother everyday communication in noise
ClearSense Audio, running on a smartphone with ordinary earphones, applies mpWAV clarity enhancement to everyday conversation and public hearing welfare.
Robotics
Robots run motors and cooling fans while delivering prompts and responses through their own speakers.
Users may call from any direction, not just the front, while nearby conversation and room reflections also enter the microphones.
mpWAV connects echo cancellation, multi-microphone noise removal, wake-word detection, localization, and recognition-plus-dialogue into the robot's complete voice interface.
mpAEC
Removes echo from the robot's own prompts and responses.
mpBeamforming·mpAB
Removes motor, fan, and nearby-conversation noise, strengthening the target voice.
mpWWD
Detects the wake word and activates the robot's listening state.
mpLocalization
Estimates the direction or position of the user's voice.
mpASR
Converts commands into text or robot control commands.
mpLLM
Interprets natural-language commands, questions, and dialogue context.
Conversational Kiosk
Accurate recognition in store noise is essential for kiosk voice ordering.
On top of that, users phrase menus and options in many ways, missing information requires follow-up questions, and the final result must reach the ordering system.
mpWAV connects a linear microphone array and multi-channel audio I/O hardware, mpAB preprocessing, mpASR recognition, and mpLLM to deliver dialogue generation and automated ordering.
Linear mic array & multi-channel audio I/O HW
Captures the user's speech in front of the kiosk on multiple channels.
mpAB
Removes store conversation, background music, and kiosk prompt echo, enhancing the user's voice.
mpASR
Converts menu names, options, quantities, and order utterances into text.
mpLLM
Interprets intent and context to understand the order, asking follow-up questions for missing information.
Automotive and Mobility
A car cabin combines engine and road noise, HVAC, music, and passenger conversation at once.
The voice interface must cancel car audio echo, recognize the wake word, identify which seat or direction spoke, and then understand commands and natural-language requests.
mpAEC
Removes echo from car audio and voice prompts.
mpBeamforming·mpAB
Removes driving noise and passenger conversation, enhancing the user's voice.
mpWWD
Detects the automotive wake word.
mpLocalization
Estimates speech direction and seat position.
mpASR
Recognizes navigation, climate, media, and other vehicle commands.
mpLLM
Connects natural-language requests and context to vehicle functions.
Home Appliances, Smart Devices and Earbuds
Earbuds, wearables, and small smart devices often cannot fit multiple microphones due to size, layout, and power constraints.
mpNC removes ambient noise from a single microphone input, making voice input improvement and a voice interface feasible on single-mic products.
mpNC
Reduces everyday ambient noise from a single microphone input and enhances the user's voice.
mpWWD
Detects the wake word to activate product functions.
mpASR
Recognizes commands and short utterances on-device.
mpLLM
Connects product functions with natural-language requests.
Applicable products
Meeting and Voice Chat
Meetings and voice chat combine multiple talkers, speaker echo, far-field voices, and overlapping conversation.
mpWAV uses mpSeparation to separate multi-party input by speaker, mpDiarization to tell who spoke when, then mpASR and mpLLM to extend into transcripts and summaries.
mpSeparation
Separates multi-party or overlapping voice input.
mpDiarization
Separates per-speaker segments and speaker turns.
mpAEC
Removes echo from remote participants played through the loudspeaker and re-entering the microphone.
mpBeamforming
Removes ambient noise and enhances the meeting audio.
mpASR
Transcribes the meeting audio.
mpLLM
Organizes summaries and key decisions.
Applicable services
Industrial Acoustic Anomaly Detection
On a production line, many motors and machines run at once, making the target equipment's sound hard to measure.
Building on its real-environment noise processing, mpWAV estimates the target equipment's acoustic signal and detects normal versus abnormal patterns.
Acoustic preprocessing
Estimates the target equipment's acoustic signal from complex line noise.
Anomaly detection model
Detects anomalies from the target equipment's acoustic signal.
Acoustic capture edge HW
Collects acoustic data directly at the equipment.
Licensing
Develops data-driven models fit to your equipment and production environment.
Applicable areas
Defense and High-Noise Environments
Defense and special environments can require handling extreme ambient noise, limited networks, and embedded compute constraints together.
mpWAV delivers an embedded system solution combining multi-channel preprocessing, wake-word detection, and on-device recognition.
mpAEC · mpBeamforming · mpAB
Multi-channel preprocessing amid extreme ambient noise.
mpWWD
Wake-word based interface activation.
mpASR
Network-independent on-device speech recognition.
Embedded system solution
Delivers a real-time embedded system solution.
Hearing Support and Smart Listening
ClearSense Audio is a smart listening solution that helps people hear conversation in ambient noise crisply and clearly, using a smartphone and ordinary earphones.
It extends mpWAV's single-microphone noise removal and voice enhancement into a range of B2C products and public hearing welfare.
mpNC
Removes ambient noise from the captured signal and strengthens the other person's voice.
Mobile processing
Supports product structures using a smartphone and ordinary earphones.
Effortless listening experience
Delivers a differentiated listening experience with the smartphone and ordinary earphones you already own — no dedicated equipment.
Application Technology Matrix
Scroll horizontally to see all columns
| Application | Input | Speech Refinement | Recognition & Dialogue | Implementation |
|---|---|---|---|---|
| Robots | mpAEC·mpBeamforming·mpAB | — | mpWWD·mpASR·mpLLM·mpLocalization | Mic Array·DSP·FPGA·AP |
| Kiosks | mpAEC·mpBeamforming·mpAB | — | mpWWD·mpASR·mpLLM | Linear Mic Array·AP·DSP |
| Vehicles & mobility | mpAEC·mpBeamforming·mpAB | — | mpWWD·mpASR·mpLLM·mpLocalization | Mic Array·DSP·FPGA·AP·SoC |
| Appliances, smart devices & earbuds | mpNC | — | mpWWD·mpASR·mpLLM | Edge·lightweight models |
| Meetings & voice chat | mpAEC·mpBeamforming·mpAB | mpSeparation·mpDiarization | mpASR·mpLLM | PC·Server·Edge |
| Factory anomaly detection | Preprocessing | — | Anomaly Detection | Mic Input System·License |
| Defense & special environments | mpAEC·mpBeamforming·mpAB | — | mpWWD·mpASR | Embedded System |
| Hearing assistance | mpNC | — | — | Mobile·ClearSense Audio |
This table is a general guide. Real projects finalize the combination based on product hardware, data, latency, and service goals.
Find the Right Technology
| Product problem | Recommended technology |
|---|---|
| Limited microphone options | mpNC |
| Echo from the product's own speaker | mpAEC |
| Removing ambient noise with multiple mics | mpBeamforming |
| Echo and noise at the same time | mpAB |
| Detecting a wake word | mpWWD |
| Separating multi-party voice input | mpSeparation |
| Knowing the speaker's direction or position | mpLocalization |
| Telling who spoke when | mpDiarization |
| Converting speech to text or commands | mpASR |
| Understanding context and intent | mpLLM |
| Multi-channel audio I/O and real-time processing | Multi-channel Audio I/O HW |
Delivery Options
Apply mpWAV technology while keeping your existing product and ASR.
Before-and-after comparison with your real product audio and field data.
Multi-channel input, speaker output, and the AEC reference connected in one structure.
Runs mpWAV technology on your embedded compute platform.
Fits mpASR and mpLLM to the target product's memory and compute budget.
Preprocessing, acoustic analysis, and conversational interfaces co-developed around your product data.
Dedicated voice interface structures for volume production and miniaturization.
Application Experience
ASR preprocessing for home, showroom, and care robots
Multi-channel audio I/O and preprocessing modules, recognition and automated ordering
ASR preprocessing in real in-vehicle noise
Motor anomaly detection validated amid complex production noise
ClearSense Audio usability reviewed at welfare centers
From Use Case to Product
Wake-up, commands, dialogue, meeting records, or anomaly detection — we scope what you need.
Microphone count, speaker output, user distance, and the noise environment.
Single/multi-channel raw audio, the AEC reference, and your current ASR results.
Select from mpNC, mpAEC, mpBeamforming, mpAB, mpWWD, mpSeparation, mpLocalization, mpDiarization, mpASR, and mpLLM.
Evaluate audio quality, detection rates, localization and diarization, recognition, dialogue results, and latency.
Apply as software, a mic array, DSP, FPGA, AP, or an edge environment.
Deliver recognition and dialogue results to robot control, ordering, vehicle functions, meeting systems, or alerting.
Verify performance where the product is actually used, under its real noise.
Review follow-up products, platforms, model optimization, and SoC expansion.
FAQ
Robots, kiosks, vehicles, home appliances, smart devices and earbuds, meetings and voice chat, factory acoustic anomaly detection, and hearing assistance.
Defense and special environments are handled as dedicated hardware and co-development areas.
mpNC — noise removal from a single microphone.
Earbuds, wearables, and small smart devices, where size, component layout, and power rule out an array.
If your product can take several microphones, mpBeamforming removes more noise.
Yes — the two work together.
mpWWD detects the wake word while mpLocalization uses multi-microphone input to estimate the direction or position of the voice.
Yes.
mpAB removes store noise and prompt echo, mpASR converts speech to text, and mpLLM interprets menu, options, quantity, and context.
The real ordering API and target hardware still need validation.
Yes — there is a demo video in the meetings and voice chat section of this page.
It shows mpSeparation splitting speech by speaker while several people talk over each other in noise.
Add mpDiarization to work out who spoke when and mpASR on top, and you get a per-speaker transcript.
Five things decide it: microphone count, whether the product has a speaker, distance to the user, how many people speak at once, and the final feature.
Earbuds have one microphone, so mpNC. A robot has both its own speaker and motor noise, so mpAB plus mpWWD and mpLocalization.
The per-industry sections above set out each combination — and if yours is not there, tell us the product conditions and we will put one together.
No.
Depending on microphone count, speakers, user position, number of talkers, and required features, you can select only what you need.
Yes.
Starting from the usage scenario and required voice features, we can review microphone structure, the technology combination, and data collection conditions.
Robots, kiosks, mobility, home appliances, smart devices, hearing assistance, meetings and voice chat, defense, and factory anomaly detection — as AI drives voice interfaces into more categories, the technology applies to nearly any field.
Design Your Application Stack
Tell us your product type, microphone count, speaker structure, noise environment, and the features you want to build — we will review the right mpWAV stack.
From single-mic and multi-channel noise-removal preprocessing to speaker, wake-word, and location analysis, on-device ASR, and a conversational LLM — we deliver a solution tailored to your product.
