Products and Platforms

The full AI voice stack, delivered in the form your product needs

mpWAV provides the complete voice interface stack — single- and multi-microphone enhancement, wake-word detection, sound source localization, speech separation, speaker diarization, on-device speech recognition, and a conversational LLM.

Depending on your microphone configuration, compute platform, development stage, and production plan, it can be applied as an SW solution, an SW/HW integrated module, FPGA·DSP·AP porting, technology licensing, or semiconductor IP·SoC collaboration.

Discuss your productRequest a PoC

From Technology to Implementation

What do we mean by products and platforms?

mpWAV's products and platforms are the software, AI models, hardware modules, and embedded implementations that package our voice algorithms for your product.

The same technology can be validated early with a PC-based library or PoC, while productization may require FPGA·DSP·AP porting, a microphone array module, or a semiconductor IP structure.

We don't just hand over technology — we shape it into something that runs inside your product.

Product Portfolio

Pick the delivery form your product needs now

Software

The full voice interface stack as software

Voice enhancement, speaker analysis, localization, wake-word detection, recognition, and dialogue — delivered as software that fits into your current system.

  • mpNC
  • mpAEC
  • mpBeamforming
  • mpAB
  • mpSeparation
  • mpDiarization
  • mpLocalization
  • mpWWD
  • mpASR
  • mpLLM

A good fit when

  • You want to validate performance on real data first
  • You want to improve the front end of your existing ASR
  • You want single- or multi-mic noise-refinement features in software
  • You are expanding meeting, kiosk, or robot service features
Discuss Software

Edge AI package

On-device speech recognition

Connects mpWWD and mpASR for wake-word detection and voice command features running locally or at the edge.

Depending on the product, mpNC, mpAEC, mpBeamforming, or mpAB can be combined as the input front end.

A good fit when

  • You want lower network dependence
  • You want voice commands processed inside the product
  • You are adding low-power wake-word detection to an edge device
  • You are adding wake-word and voice command features to a smart device
Discuss on-device AI

Microphone array module

Multi-channel input and preprocessing in one structure

Connects multiple MEMS microphones, speaker output, the AEC reference, and speech preprocessing.

A good fit when

  • You need to handle far-field user speech
  • You need recognition in noisy places or speaker direction finding
  • You need to handle the product's speaker echo
  • You want to validate software and hardware together
See the mic array module

FPGA·DSP·AP porting

Real-time execution on your compute platform

We optimize the validated technology for your DSP, FPGA, or AP — multi-channel audio input, memory, compute, and latency.

A good fit when

  • You are moving a PC PoC into the real product
  • You need real-time multi-channel processing
  • You must keep your existing DSP·AP platform
  • You need to connect the voice stack to ASR and product APIs
See FPGA·DSP·AP porting

Licensing·co-development

Technology collaboration around your data and goals

mpWAV's preprocessing, refinement, recognition, conversational AI, or industrial acoustic analysis — licensed or co-developed for your product.

A good fit when

  • You are applying the technology to a product line long-term
  • You need optimization for specific industry data
  • You are developing hardware and software together
  • You need custom analysis such as factory anomaly detection
Discuss licensing·co-development

Semiconductor IP·SoC

Dedicated structures for volume and miniaturization

Building on FPGA·DSP implementation experience, we review semiconductor IP or dedicated SoC collaboration that accounts for the power, size, and cost of volume products.

A good fit when

  • You need cost and power optimization for volume production
  • You need a dedicated chip structure for miniaturization
  • You are a semiconductor company exploring IP collaboration
Discuss SoC·semiconductor partnership

Software

Select the voice module that matches your product's problem

The software portfolio spans three areas: improving voice input, refinement, and recognition with dialogue.

Apply a single technology, or combine several into one pipeline.

Speech Enhancement

Improving voice input

mpNC

Removes ambient noise from a single microphone input.

Fits earbuds and small smart devices with limited microphone options.

mpAEC

Removes echo from the device's speaker re-entering the microphone.

mpBeamforming

Uses multiple microphones to remove ambient noise and strengthen the target voice.

mpAB

Integrates mpAEC and mpBeamforming to handle echo and noise together.

Speech Refinement

Speech separation and speaker diarization

mpSeparation

Separates multi-party or overlapping voice input into per-speaker voices.

mpDiarization

Separates who spoke when in multi-party audio.

Speech Recognition & Dialogue

Wake-word detection, recognition, and dialogue understanding

mpWWD

Detects the wake word and activates the product's listening state.

mpASR

On-device end-to-end recognition converting speech to text or commands.

mpLLM

A conversational language model connecting intent, context, and product functions.

mpLocalization

Estimates speech direction or position from multi-microphone input.

See the technology overview

Modular or Integrated

Pick only what you need — or integrate the whole stack

Front end for your existing ASR

Microphone → mpNC / mpAEC / mpBeamforming / mpAB → Existing ASR

Improves the input signal while keeping the recognition engine you already use.

Robot voice interface

mpAB → mpWWD → mpLocalization → mpASR → mpLLM → Robot Command

Connects preprocessing, wake-up, localization, recognition, and dialogue.

Kiosk voice ordering

Multi-channel Audio I/O HW → mpAB → mpASR → mpLLM → Ordering API

Connects store-noise preprocessing, menu recognition, and conversational ordering.

Single-mic smart device

Single Mic → mpWWD → mpNC → mpASR → mpLLM → Device Function

A single-microphone stack for products with limited microphone options.

Meetings & voice chat

mpAEC → mpSeparation → mpDiarization → mpASR → mpLLM → Transcript / Summary

Connects speech separation, speaker diarization, transcripts, and summaries.

On-Device Edge AI

From wake word to recognition — inside the product

A product voice interface needs reliable wake-word detection and accurate recognition even in noisy places.

mpWAV builds this around mpWWD, mpAB, and mpASR running inside the product.

Wake word

mpWWD

An always-on product detects the predefined wake word and starts recognition.

Input enhancement

mpNC·mpAEC·mpBeamforming·mpAB

Refines the input to match the product's microphone and speaker configuration.

Recognition

mpASR

Converts commands, menu names, and natural speech to text.

Typical applications

  • Conversational kiosks
  • Service robots
  • Vehicle voice interfaces
  • Smart home·appliances
  • Small smart devices
  • Network-constrained edge products

On-device feasibility depends on model size, precision, memory, CPU·GPU·NPU, and the product's response-time requirements.

Discuss on-device AI

Microphone Array Module

From multi-channel capture to localization and preprocessing

Multi-microphone products must connect mic count and layout, channel synchronization, speaker output, and the AEC reference in one hardware structure.

mpWAV has multi-channel audio hardware design experience including a robot·kiosk linear array and a meeting circular array.

mpBeamforming

Ambient noise removal using multi-microphone spatial information

mpLocalization

Speech direction or user position estimation

mpAEC

Echo cancellation using the speaker reference

mpAB

Integrated echo and noise processing

mpASR

Converting the refined audio to text and commands

Examples

Robot·kiosk linear array

  • 2–6 channel MEMS microphones, 16 kHz·16-bit input
  • 48 kHz·16-bit stereo reference input and digital amplifier output

Consultation·meeting circular array

  • 6-channel MEMS microphones, 16 kHz·16-bit input
  • 48 kHz·16-bit stereo reference input and digital amplifier output

Ask about the mic array module
Two circular MEMS microphone array boards, a multi-channel kiosk audio board with linear microphone arrays, and the boards installed in an actual kiosk
Multi-channel microphone array boards and a kiosk deployment

Embedded Processing Platforms

From software to the real product platform

Running validated mpWAV technology on a real product requires porting that accounts for channel count, memory, latency, power, and audio interfaces.

mpWAV delivers the optimal solution for your real product environment.

DSP∙FPGA∙MCU

Low-latency processing inside the product

Optimizes compute, memory, and processing time for the product's processor and environment to secure real-time operation.

AP

OS and service integration

Connects audio input, existing ASR integration, product API integration, and applications in one software structure.

Edge AI Accelerator

mpAB·mpASR·mpLLM execution review

Reviews model size, execution paths, and end-to-end pipeline latency matched to the product's CPU, GPU, or NPU.

Ask about FPGA·DSP·AP porting

Solution Packages

Technology packages composed for your product

Robot Voice Interface Package

From wake-up to dialogue

Mic Array + mpAB + mpWWD + mpLocalization + mpASR + mpLLM + DSP / AP

Problems solved

  • Wake-word detection
  • User direction estimation
  • Motor·fan noise
  • Robot speaker echo
  • Natural-language commands and dialogue

Conversational Kiosk Package

Conversational ordering in store noise

Multi-channel Audio I/O HW + mpAB + mpASR + mpLLM + Ordering API

Problems solved

  • Far-field speech
  • Store noise
  • Prompt echo
  • Menu·option·quantity recognition
  • Automated ordering flow

Smart Device·Earbuds Package

Voice AI for single-mic products

Single Mic + mpWWD + mpNC + mpASR + mpLLM

Problems solved

  • Structures with limited microphone options
  • Limited size and compute
  • Everyday ambient noise
  • Wake word and voice commands

Meeting Intelligence Package

Multi-party records and speaker separation

mpAEC + mpSeparation + mpDiarization + mpASR + mpLLM

Problems solved

  • Multi-party or overlapping speech
  • Speaker separation
  • Speaker echo
  • Meeting minutes
  • Summaries and structured notes

Automotive Voice Package

Natural-language vehicle interfaces in driving noise

Cabin Mic Array + mpAB + mpWWD + mpLocalization + mpASR + mpLLM + DSP / AP / SoC

Problems solved

  • Car audio echo
  • Engine·road noise
  • Per-seat speech
  • Natural-language vehicle commands

Industrial Acoustic Package

Factory anomaly detection

Mic Array + mpAEC · mpBeamforming · mpAB + Anomaly Detection + License / Co-Development

Choose by Development Stage

The right product form depends on your development stage

StageGoalRecommended product·platform
PlanningFeasibility reviewConsultation·architecture design
Data collectionAnalyzing the current problemSoftware Evaluation·USB Audio Interface
PoCBefore/after validationSoftware·AI Model
PrototypeVerifying real product behaviorMic Array·FPGA·AP Integration
IntegrationRunning on the target platformFPGA·DSP·AP Porting
Pre-productionCost·power·update reviewModule·License
ScaleMiniaturization and cost optimizationSemiconductor IP·Dedicated SoC

Early on, a software PoC identifies the technology you need; module and porting scope are best decided after hardware and the target platform are fixed.

Licensing and Co-Development

Technology scoped to your product and data

One software package cannot fit every product.

When your microphones, speakers, compute platform, environment, and service goals are unique, mpWAV technology can be licensed or co-developed for your product.

Algorithm licensing

mpNC, mpAEC, mpBeamforming, mpAB, and acoustic analysis

AI model application

mpWWD, mpASR, and mpLLM optimized per product and domain

Speech refinement technology

mpSeparation and mpDiarization connected to your systems

Hardware co-development

Microphone arrays, AEC reference, and amplifier output integrated on one board

Industry co-development

Systems built for robots, kiosks, vehicles, appliances, meetings, and factories

Follow-on expansion

Technology proven in the first product extended to other platforms and lines

Discuss licensing·co-development

Semiconductor IP and SoC

Extending the voice stack into dedicated silicon

For vehicles, appliances, smart devices, and volume products, size, power, and cost matter as much as processing performance.

Building on software algorithms, real-time FPGA implementation, and multi-channel hardware experience, mpWAV reviews voice interface IP·SoC collaboration with semiconductor companies and volume customers.

Technology blocks under review

  • mpAEC IP
  • mpBeamforming IP
  • mpAB integrated block
  • mpNC lightweight enhancement
  • mpWWD wake-word detection
  • Multi-channel Audio I/O
  • On-device mpASR
  • Per-product AI acceleration

Fields under review

  • Vehicles·mobility
  • Smart home·appliances
  • Robots·kiosks
  • Consumer smart devices
  • Defense·special environments
Discuss semiconductor IP·SoC partnership

Technology to Product Map

Technology and delivery forms matched to each application

ApplicationKey technologiesRecommended product·platform
Robots
  • mpWWD
  • mpLocalization
  • mpAB
  • mpASR
  • mpLLM
Mic Array·Software·AP·DSP·FPGA
Kiosks
  • mpAB
  • mpASR
  • mpLLM
Linear Mic Array·Edge AI·AP
Vehicles
  • mpWWD
  • mpLocalization
  • mpAB
  • mpASR
DSP·AP·FPGA·SoC
Appliances·smart devices·earbuds
  • mpWWD
  • mpNC
  • mpASR
  • mpLLM
Software·Edge AI·AP
Meetings·voice chat
  • mpAEC
  • mpSeparation
  • mpDiarization
  • mpASR
  • mpLLM
Software·PC·Server·Edge
Factory anomaly detection
  • mpAB
  • Anomaly Detection
Edge System·License
Defense·special environments
  • mpAB
  • mpWWD
  • mpASR
FPGA·DSP·SoC
Hearing assistance
  • mpNC
Mobile·ClearSense Audio

This table is a general guide. Real combinations are decided from your product's data and target platform.

Product Implementation Experience

Beyond validation — connected to products and sites

Robots

ASR preprocessing validated for home, showroom, and care robots

Kiosks

Multi-channel audio I/O and preprocessing modules, recognition and automated ordering

Mobility

Voice interfaces in real in-vehicle noise

Factory anomaly detection

Motor anomaly analysis validated amid production noise

Hearing assistance

Mobile and earphone voice enhancement via ClearSense Audio

Embedded implementation

DSP∙FPGA∙AP porting for real-time processing

See case studies

From Requirement to Product

Select the technology, then shape it to run in your product

  1. 1

    Confirm requirements

    Product type, users, voice features, and current problems.

  2. 2

    Analyze mic·speaker structure

    Microphone count, speaker output, user distance, and the noise environment.

  3. 3

    Evaluate data

    Single/multi-channel raw audio, the AEC reference, and existing results.

  4. 4

    Compose the stack

    Select what you need from the full mpWAV portfolio.

  5. 5

    Software PoC

    Validate processing results and product features in a reference environment.

  6. 6

    Decide the product form

    Library, Edge AI, mic array module, porting, or licensing.

  7. 7

    Integrate the platform

    Apply to DSP, FPGA, AP, or an edge AI environment.

  8. 8

    Connect product APIs

    Deliver results to robot control, ordering, vehicle functions, minutes, or alerts.

  9. 9

    Field verification

    Verify on the real product, in its real environment.

  10. 10

    Production & expansion

    Modules, licensing, follow-on products, and dedicated SoC structures.

Discuss products & platforms

Project Information

Share the following and we can review the right configuration

You do not need every item settled to start a conversation. Begin with the voice problem your product hits most often and the feature you are aiming for.

Product

  • Product or service type
  • Application field
  • Primary users
  • Usage locations
  • Development stage
  • Launch schedule
  • Expected volume

Acoustic structure

  • Microphone count
  • Microphone positions
  • Speaker count
  • AEC reference availability
  • User distance
  • Noise environment
  • Multi-party speech

Required features

  • Single-mic noise removal
  • Speaker echo cancellation
  • Multi-mic noise removal
  • Wake-word detection
  • Direction estimation
  • Speech separation
  • Speaker separation
  • Speech recognition
  • Natural-language dialogue
  • Anomaly detection

System environment

  • CPU
  • DSP·MCU
  • FPGA
  • AP
  • GPU·NPU
  • Operating system
  • Memory
  • Latency budget
  • Power budget

Data

  • Raw audio
  • Per-channel microphone data
  • Speaker reference
  • Ground-truth text
  • Existing ASR results
  • Failure cases
  • Menu·command·domain data

FAQ

Frequently asked questions about mpWAV products & platforms

Six delivery forms — software, on-device AI, microphone array modules, FPGA·DSP·AP porting, technology licensing, and semiconductor IP·SoC collaboration.

Which form suits you depends on where your product is. Software for algorithm validation; porting or modules once you are heading to production.

No — you take only what you need.

Technologies come as individual modules, or we connect several into one integrated configuration.

Delivery units and licensing are agreed per project.

Yes. mpNC is built for single-microphone input.

It exists for earbuds and small smart devices that cannot fit an array.

Yes — whichever delivery form you choose.

Software, a microphone array module, or a ported build: mpNC, mpAEC, mpBeamforming, and mpAB all sit in front of recognition.

The only case where the engine changes is if you decide to take mpASR and move recognition to us as well.

A linear microphone array with multi-channel audio I/O hardware, plus mpAB, mpASR, and mpLLM.

Connect your ordering system API and it removes store noise and prompt echo, then interprets menu, options, and quantity before handing them over.

Yes.

mpSeparation splits overlapping speech by speaker, mpDiarization works out who spoke when, and mpASR turns each speaker's audio into text.

Yes — mpLLM, fitted to your usage scenario.

Whether it runs inside your product is a separate question, settled against model version, precision, memory, accelerators, and your target response time.

Not off the shelf.

Supply form, specifications, volumes, and lead times are set per project, so contact us and we will work them out with you.

No — not as a standard chip yet.

A dedicated voice interface chip and SoC are an expansion direction we are pursuing; today we collaborate through semiconductor IP and co-development.

Start with a software PoC based on your microphone count, speaker structure, required features, target platform, and stage — then choose the right delivery form.

Build the Right Product Stack

Design the voice technology and the delivery form together

Tell us your product type, mic·speaker structure, noise environment, required features, and target platform — we will review the right mpWAV technology and delivery form.

From single-mic noise control to multi-channel echo and noise processing, speaker analysis, localization and wake-word detection, on-device ASR and conversational LLM, embedded porting and SoC expansion — matched to your development stage.

Discuss products & platformsRequest a PoC