Voice recognition modules — offline voice control

Haiben has supplied electronic components from Shenzhen since 2013, and voice is one of the areas we cover in depth. The Voice Recognition & Audio Chips & Modules group brings together voice ICs and audio ICs alongside voice modules and audio modules, plus printed circuit board assembly for voice PCBA. The emphasis is on offline voice recognition: voice control that runs locally on the device, with no internet connection required.

Offline recognition: the core idea

An offline voice recognition module — often written as a speech recognition module — does its pattern matching on the board itself. The microphone signal is conditioned, converted to digital, and compared against command patterns held in the device's own memory. There is no round trip to a cloud service, so recognition keeps working when there is no network, and response time does not depend on connection quality.

The trade-off is scope. Offline engines are built around a defined command set, so they suit products where the user speaks a small, predictable vocabulary: switch on, heat, change level, start, stop. That is exactly the profile of most small appliances and lighting products, which is why the approach has become common in them. Whether a design needs a discrete voice IC or a complete board is a decision that shapes the rest of the project, so it is worth settling early.

What a voice recognition module contains

A module packages the recognition function so it can be designed in without building the audio front end from scratch. The typical contents are:

• A microphone input with conditioning circuitry — bias, filtering and gain — so the captured signal is clean before conversion.
• An analogue-to-digital converter and a processing core running the recognition firmware.
• Memory holding the command patterns, plus the interface logic that presents results to the host — GPIO triggers, a serial bus, or both.
• Supporting passives, a clock source and a regulator, all on one small board.

On the output side, many designs pair recognition with an amplifier so the product can confirm commands audibly. Our voice ICs and audio ICs catalogue includes the VC2033 mono audio amplifier IC: operating voltage 2.5 V to 5.5 V, Class AB/Class D switching, two anti-clipping options with an anti-clipping compression range of −10 dB, and Class D output power of 3.1 W at VDD 5.0 V into 4 Ω and 1.8 W into 8 Ω (THD+N 10%), in a DFN2x2-8L package. The sibling part VC8210 is a single-channel Class AB/D audio amplifier IC in the same DFN2x2-8L package, rated 3.1 W in Class D.

Where offline voice control is designed in

Offline control fits products where the user's hands are busy or the controls are awkward to reach. Our MIC store carries a dedicated Voice Recognition PCBA Board group with many offline-voice PCBA listings, and the published examples show the range of applications: an AC42228-SH001 offline smart voice recognition kettle circuit board PCBA; a VC100GA-V1.0 tea bar module PCBA with ASR voice control; an ASR offline voice control night light PCBA module; a mosquito repellent lamp voice PCBA; and smart home voice control module PCBA boards.

These sit in different product families — kettles, tea bars, night lights, fans and smart-home boards — but share the same design logic: a fixed command vocabulary, local processing, and a simple interface to the appliance's existing control board. Buyers specifying voice modules for this kind of product usually start from the command list and the mechanical fit, then work back to the electronics.

Behind the catalogue, the Weforsemi line is operated by our wholly-owned subsidiary, and the group has an integrated chain covering R&D, wafer testing, packaging and AI voice chip production and sales. That chain matters when a design moves from a module on a bench to a board in a production run: the same organisation follows the part from development through packaging.

Specifying a module or a voice PCBA

There are two practical routes. A standard module is the quicker option when its command count, interface and supply voltage suit the product, and it can be treated as a component on the customer's own board. A voice PCBA goes further: Haiben assembles the recognition circuit together with the product's control functions on one board, so the appliance maker integrates a tested sub-assembly rather than a bare module.

For either route, the following information drives the choice:

• The wake word and command set, and whether speakers must be supported or the vocabulary fixed at build time.
• Supply voltage and current budget, taken from the host product's existing rail.
• The interface to the host — number of trigger outputs, serial protocol, or both.
• Acoustic requirements: microphone placement, expected distance to the user, and the noise level in normal use.
• Board outline, connector positions and mounting, agreed against the enclosure drawing.
• Audio output, if spoken feedback is required; this is where an amplifier such as the VC2033 or VC8210 is selected.

Our PCB assembly service covers the voice PCBA route, from board layout through assembly. Because the voice function, the amplifier and the host control circuit are specified together, interface questions are resolved before tooling rather than after first articles.

Frequently asked questions

Does an offline voice recognition module need an internet connection?

No. Recognition runs locally: the microphone signal is processed on the board and matched against stored command patterns, so the product works without a network. The command vocabulary is defined during development rather than drawn from an online service.

What is the difference between a voice module and a voice PCBA?

A voice module provides the recognition function itself, ready to be designed onto a host board. A voice PCBA integrates that function with the product's control circuitry on one assembled board — our published examples include kettle, tea bar, night light, fan and smart-home boards — so the appliance maker receives a tested sub-assembly.

Can Haiben support both the recognition part and the audio amplifier?

Yes. The catalogue groups voice ICs and audio ICs together with voice and audio modules, so the recognition device and the amplifier can be specified as one exercise. For voice recognition chip questions, start from the voice ICs page, which lists the parts and their published specifications.