Noise reduction offline speech recognition module

Main Product Parameters:

Dimensions: 30 x 20 x 2.2 mm

Package: SMD56

Peripheral Interfaces: UART/GPIO/PWM/I²C/ADC

Memory/Flash: 8MB PSRAM / 1MB SRAM / 8MB NOR Flash

Microphone Array: 2-channel analog microphones

Signal Array: 2 reference signals (supports voice interruption function)

Amplifier: Class D, 8Ω / 0.75W output

Wake Words: 3 sets (customizable)

Command Words: 200 sets (customizable)

Product Description

The dual-microphone offline voice module is a voice interaction solution designed for home environments, enabling voice control that is more direct, natural, and convenient compared to traditional manual controls (machine, remote, smartphone). This enhances the user experience of products. The voice module boasts a wealth of peripheral resources, including UART, I2C, SPI, PWM, and ADC.

The module employs acoustic front-end solutions such as voice enhancement, sound source localization, and echo cancellation, along with features like voice wake-up and offline recognition. It uses the latest generation microphone array algorithm, integrating traditional signal processing with neural network capabilities to significantly improve front-end acoustic processing.

This module can be widely applied to smart home appliances, 86 boxes, toys, lighting, industrial, medical, automotive, security, and lighting products that require voice control.

Product Features:

▶ Purely offline solution, capable of voice control even without an internet connection.

▶ Utilizes FreeRTOS for fast system startup upon power-on.

▶ Meets the voice interaction needs of home appliances:

a) Dual-microphone front-end processing algorithm, catering to most noise reduction requirements;

b) Echo cancellation, supporting interruptive announcements;

c) New generation neural network wake-up and recognition technology;

d) Supports built-in audio playback.

▶ Superior low false wake-up rate compared to industry standards.

▶ High recognition rate for Mandarin with regional accents.

Function Introduction：

Far-field Voice Pickup
Capable of 360-degree far-field voice pickup up to 5 meters. Equipped with automatic gain control, it adapts to the user’s volume to ensure consistent audio quality after noise reduction.

Environmental Noise Reduction
Widely applicable for noise reduction in home, car, office, and other environments, while preserving maximum voice clarity.

Echo Cancellation
Supports user interactions where the device can be interrupted by the user during content or music playback, allowing for a more natural interaction experience.

Beamforming
Enhances voice capture by focusing on sounds within the beam range where the voice originates, suppressing external noise.

Voice Broadcasting
The device provides voice responses when the user wakes it up and issues commands, or proactively offers prompts. The purpose is to provide feedback to the user through voice responses upon receiving voice commands or in suitable scenarios.

Offline Commands
In the wake-up state, the user can issue commands within a specified range, which the voice module processes or forwards to the host for further action.

Peripheral Communication
The chip processes microphone inputs and communicates with other devices via USB, I2S, SDIO, and other interfaces, or through general-purpose programmable I/O.

Electrical Control MCU
The chip offers memory and standard serial protocols, allowing developers to create simple electrical control logic for products (e.g., fans, heaters, humidifiers), effectively replacing and acting as the original product’s control MCU.