Audio Front-End for real-world voice AI
In front of your ASR, mpWAV cuts echo and noise while preserving the speech recognition needs. Drop it into your existing product — no cloud, no dedicated NPU.
We compare recognition before and after, on your recordings and your environment.
Before / after
Echo and noise come down; the speech your recognizer needs stays in.
mpWAV Solution DemoHeadphones or a speaker make the difference easier to hear.
BEFORE
Raw signal from the microphones
mpWAV
Echo and noise removal (mpAB)
AFTER
Signal handed to the speech engine
mpWAV does not replace your speech engine. It changes what the engine receives.
Send us a note and a research engineer replies within 2-3 business days.
Why audio preprocessing?
Speech recognition that works in a quiet office loses accuracy once the product ships into a real environment.
The reason is that the input speech is already damaged by noise and echo before it ever reaches the ASR.
01
People talking nearby and the background noise of a store or office arrive together with the user's voice.
02
Sound from a TV or the product's own speaker loops back into the microphone.
03
Motors, fans, and moving parts interfere with recognition.
04
Driving speed and road surface keep changing the character of the noise.
05
The farther the user moves away, the quieter the voice and the louder the noise.
06
Reflected speech mixes with the original and distorts the input signal.
What you need is not just a better ASR. You need an Audio Front-End in front of it.
Why mpWAV
In front of your speech engine, mpWAV turns sound from a real environment into a signal that is easier to recognize.
Instead of swapping in a new speech engine, it gives the ASR you already run a more stable input.
01
Ordinary noise removal can sound clean to a person while damaging the very features speech recognition depends on.
mpWAV focuses on reducing interference while preserving as much of the target speech as recognition needs.
Speech Preservation
02
There is no need to replace your whole system with a different speech engine.
mpWAV sits in front of your existing ASR and improves the signal that engine receives.
Keep Your Existing ASR
03
It does not assume a cloud connection or a dedicated AI accelerator, and it is applied with the compute and memory budget of an embedded product in mind.
You can evaluate whether it fits without significantly changing your current product architecture.
On-device
04
Microphone count, placement, user distance, room structure, and noise characteristics differ for every product.
mpWAV works out the Audio Front-End configuration your real usage environment calls for.
Environment-specific Integration
Verified performance
The core mpWAV preprocessing has been verified in accredited testing.
The figures below come from the TTA test results for mpAB.
42.088 dB
Stereo echo and noise suppression
64 MHz
Clock frequency
0.571 MB
SRAM required
0.852
Real-Time Factor
Measured on mpAB in accredited TTA testing
Measured in a TTA test report (2023, Ministry of SMEs and Startups Didimdol programme) with mpAB as the article under test.
Proven in real environments
We validate whether the preprocessing holds up in real product conditions, from care robots to far-field speech.
ROBOT · HOME ENVIRONMENT
In homes where a TV and everyday household noise run together, mpWAV preprocessing was applied so the robot receives the user's wake word and commands reliably.
ROBOCARE Cami-friends
HUMANOID · LIVE DEMO
At ETRI Conference 2025 we demonstrated voice input and system behaviour through a real robot interface.
ETRI
We compare recognition before and after, on your recordings and your environment.
Applications
The mpWAV Audio Front-End is built for products that have to understand a human voice — robots, vehicles, kiosks, and smart devices.
The same acoustic signal processing extends to anomaly detection, defence and security, meetings, and hearing assistance.
ROBOT
Service, home, care, and guide robots have to handle not only ambient noise but the motor and actuator noise they generate themselves.
mpWAV keeps the wake word and the spoken command reaching the engine even in those conditions.
Motor Noise · Far-field Speech
AUTOMOTIVE & MOBILITY
Road noise, in-car audio, and passengers talking all arrive at once, and the target voice still has to get through.
Road Noise · Echo · Multiple Speakers
KIOSK
In stores, hospitals, and public offices, a barrier-free voice interface has to take questions and commands over the background noise of the room.
Background Noise · Distance
SMART DEVICE & APPLIANCE
In the homes where TVs, appliances, and smart-home devices actually run, conversation, household noise, and speaker echo happen at the same time.
mpWAV keeps the voice interface working in those conditions.
TV Audio · Household Noise · Echo
INDUSTRIAL
Detecting abnormal motor and equipment sounds acoustically on a noisy production line.
Acoustic Monitoring
DEFENSE & SECURITY
Long-range signal detection and multimodal recognition, applied in national R&D projects for environments with extreme noise.
Long-range Detection · Multimodal Recognition
MEETING & VOICE CHAT
Speech separation and speaker diarization for meeting minutes, online meetings, and voice chat.
Speech Separation · Speaker Diarization
HEARING ASSISTANCE
ClearSense Audio raises speech clarity in noise on ordinary earbuds and phones.
Speech Enhancement
The product changes.The acoustic problem repeats.
Noise · Echo · Distance · Reverberation · Source separation
Tell us the product and where it is used, and we will look at whether it fits.
Audio Front-End technology
Echo, ambient noise, far-field speech, and a product's own noise are hard to solve with a single algorithm.
mpWAV combines AEC · Beamforming · Noise Cancellation · Adaptive Processing to match the microphone layout and the environment of the real product.
Microphone input
mpWAV Audio Front-End
Existing ASR
Acoustic Echo Cancellation
Sound played by the product's own speaker loops back into the microphone, mixes with the user's voice, and can get in the way of recognition.
mpAEC reduces that acoustic echo so the engine receives a steadier input signal.
Speaker Echo
Beamforming
Using multiple microphone inputs, it captures speech from the target direction and raises input quality when the speaker is far away.
Far-field Speech · Directionality
Noise Cancellation
Household noise, road noise, and the product's own motor and fan noise all interfere with recognition, and mpNC reduces them.
The goal is not simply sound that is pleasant to a listener, but an input signal a recognizer can use.
Environmental Noise · Self Noise
Adaptive Audio Front-End
Working from the actual input signal, mpAB reduces echo and noise while preserving as much of the target speech as recognition needs.
Adaptive Processing · Speech Preservation
We do not sell one algorithm.We build the Audio Front-End the product needs.
Microphone count, placement, user distance, and noise characteristics decide which combination of techniques applies and how it is integrated.
Integration
mpWAV is not a technology that stops at the PoC.
Depending on your processor, microphone layout, and stage of development, we can look at anything from a software library to hardware porting, an audio module, or production.
SOFTWARE LIBRARY
We assess applying the mpWAV preprocessing as a software library within your product's existing processor environment.
That means checking how it integrates with your current audio and recognition pipeline without a major change to the product architecture.
Minimum Hardware Change
PORTING
For DSP, FPGA, or AP environments already in the product, we assess whether the preprocessing fits and what resources it needs.
Hardware-specific Integration
AUDIO MODULE
For an early PoC or during development, a module that already includes multi-channel audio I/O and preprocessing can be the faster path.
How ROBOCARE is supplied
Fast Prototyping
PRODUCTION
Once performance is confirmed through a PoC and product integration, we agree the licensing that fits your stage of development.
Where volume production and product expansion call for it, dedicated chips and semiconductor IP cooperation are considered too.
From PoC to Production
Before deciding how to integrate,find out whether it works in your environment.
We look at your product environment first, then propose how to validate and integrate.
From test to production
Start with the one environment that matters most, and compare mpWAV before and after on real data.
We look at the product and the problem first, then work through only the scope the validation needs.
We go over the product, the number and placement of microphones, user distance, the main noise environment, and what is going wrong today.
What we need to start
Product type · microphone count · user distance · main noise environment
If you have speech recorded in the real environment, we analyse the current problem and the signal characteristics first.
If there is no recording yet, we can define the test environment and recording conditions together.
We compare the same conditions with and without mpWAV to see the effect of preprocessing and whether it fits the product.
Recognition results, latency, and compute load — we define the KPIs the project actually needs together.
Once the PoC confirms the effect, we decide how to integrate it with your processor and your audio and recognition pipeline.
The integrated product does not stay in lab conditions — we check it again where it is actually used.
Distance, room, the product's own noise, and real usage patterns all get reviewed on site.
A validated configuration goes into the product, and can extend to follow-up models or a product line where that is wanted.
Licensing, technical support, and how it reaches production are agreed per project.
You can start without any recordings.
Product type, microphone count, user distance, and the main noise environment are enough for us to work out how a validation could run.
Tell us about the product environment and we will start from the validation step that fits.
FAQ
It is the stage that removes noise and echo before the signal reaches your speech recognition engine.
It raises the quality of the input signal without changing the recognition engine itself, so it applies to a speech system you have already built. mpWAV delivers echo cancellation, beamforming, and noise suppression as a single on-device module.
No. You keep the engine you have.
mpWAV is a preprocessing layer that sits between the microphone input and the recognition engine. It applies the same way to a commercial API or an in-house engine, with no retraining.
It runs in real time at 64 MHz clock, 0.571 MB of SRAM, and a 543 KB binary.
The real-time factor is 0.852, and it needs no cloud connection and no dedicated NPU. It drops into embedded environments such as robots, kiosks, and vehicles.
It was verified in an accredited third-party test.
A TTA test report from the SMBA Didimdol program recorded 42.088 dB of noise reduction and a real-time factor of 0.852 (65 dB speech, 60 dB noise). mpWAV also holds two CES 2024 Innovation Awards and NET new-technology certification from the Ministry of Trade, Industry and Energy.
You can choose a software library, a multi-channel audio I/O preprocessing module, a DSP or FPGA port, a technology license, or an SoC partnership.
Which one fits depends on your development stage and compute platform.
It starts with a PoC on your own recorded speech data.
The sequence is environment analysis, data review, PoC and performance validation, product integration, on-site tuning, and then scaling to production. We can begin with the environment analysis even if you have no field recordings yet.
Tell us your product type, audio I/O configuration, usage environment, and current errors — we will recommend the right validation method and integration structure.
You don't have to commit to a large project up front. Start a PoC with your most important environment and data.