Array Microphone Solution

AI Smart Voice Solution

Product Direction

● In voice interaction, the process generally consists of five stages: target signal acquisition, voice enhancement, speech recognition, semantic understanding, and information feedback. Among these, speech recognition technology is already quite mature. Our product's key advantage lies primarily in voice enhancement. We provide signal acquisition and voice enhancement technologies for voice interaction applications;

● Voice enhancement technology as the core application of this product

√ Beamforming

√ Sound source localization

√ Directional audio pickup

√ Sound source separation

√ Beamforming

√ Echo cancellation

√ Noise reduction, dereverberation, etc.

AI Smart Voice Solution: Mainboard Overview

◆ The product adopts a core board + baseboard architecture, enabling configurable performance based on requirements;

◆ The core board features TI’s AM57xx ARM+DSP multi-core high-performance processor;

◆ ARM cores handle business logic, while DSP cores process audio and video data analysis;

◆ DSP application development SDK enables easy implementation of audio and video data analysis algorithms;

◆ Multiple McASP audio interfaces + CSI/USB camera video input;

◆ Suitable for audio and video processing, smart education, smart robots, and other AI application fields

AI Smart Voice Solution: Array Microphone Board Overview

Circular array microphone

Linear array microphone

AI Smart Voice Solution: Core Board Overview

SOM5728 Core Board

2*ARM Cortex-A15 + 2*C66x DSP

AI Smart Voice Solution: Product Advantages

Acoustic Specifications

Microphone Array Type

Circular Array

Linear Array

Number of Microphones

13-mic (6+6+1), 7-mic (6+1), 8-mic (4+4)

8, 4

Beam Direction

360°

Effective Pickup Distance

0~10 meters

0~20 meters

Effective Beam Width (3dB bandwidth)

70°

Sound Source Localization Resolution

≤ 20°

≤ 30°

Number of Separable Sound Sources

12 groups

2 groups

Audio Quality

360° high fidelity

High fidelity

Speech Recognition Rate (5m, noisy environment)

≥95%

Echo Cancellation

≥ 40 dB

Noise Suppression

Signal-to-noise ratio improvement ≥ 10 dB

AI Application Product: AI5728 Product Advantages

Sound Source Localization: Accurate sound source positioning enables a wide range of scenario-based applications.

Directional Audio Pickup: Directional pickup technology effectively suppresses noise outside the target range, improving voice clarity.

Beamforming: The core algorithm of microphone arrays, beamforming enhances sound directivity in the target direction.

Sound Source Separation: Separates audio into multiple streams based on individual speakers, enabling broader application scenarios.

Echo Cancellation: In scenarios involving simultaneous playback and recording, this technology cancels out speaker output, capturing only the user’s voice.

Dereverberation: Dereverberation algorithms effectively reduce reverberation in speech, resulting in cleaner audio output.

Noise Suppression: Advanced noise reduction algorithms intelligently attenuate captured noise, delivering clearer audio quality.

AI Application Product: AI57xx Product Application Scenarios

Introduction to AM57xx SOM Core Board

◆ SOM572x is an industrial-grade core board developed by Xinming Technology based on TI's AM5728 processor.

◆ Industry's smallest form factor: 69mm*52mm

◆ Uses high-quality, reliable 4*80-pin precision BTB connectors to ensure signal integrity and easy installation/removal

◆ Dual Gigabit LAN, supports EtherCAT master/slave, ProfiNet, and other industrial Ethernet protocols

◆ Supports up to 10 video inputs

◆ Supports 2 CAN interfaces

◆ Supports 10 UART interfaces

◆ Supports 1 USB 3.0 & 1 USB 2.0

◆ Supports 3 LCD interfaces

◆ Supports 1 HDMI interface

◆ Supports 8 McASP audio interfaces

◆ 1 SATA interface

◆ 1 PCIe 3.0 interface, supports 1*2-lanes or 2*1-lane

◆ Supports Linux 4.9.28, DSP RTOS, and real-time RT-Linux systems

◆ Operating temperature range: -40℃~85℃ (industrial grade)

Introduction to AM57xx Processor

AM5708 Cortex®-A15 Based Processors

Benefits

● Single-core version of AM57x

● 3,500 ARM Cortex-A15 DMIPS

Sample Applications

● Human Machine Interface (HMI)

● Medical Imaging

● Test & Measurement

Key Features

● 1.0GHz* ARM Cortex-A15, Dual 213MHz ARM Cortex-M4; 750MHz** C66x DSP

● 1080p Video Decode/Encode

● 3D and 2D Graphics Acceleration

● Multiple Video Input Ports

● Display Subsystem

● Quad core Programmable Real-time Unit (PRU)/Industrial Communications Subsystem (2x dual core subsystems)

● USB3, PCIe

Power Estimates

● Typical Power: 1.5-3W (nominal voltage)

● Suspended Mode: <800mW

Package: 17x17mm, 0.65mm VCA

Availability:

● Samples: now

● Production: now

*Nominal voltage max freq. for Cortex-A15 is 1 GHz

**Nominal voltage max freq. for C66x DSP is 600 MHz

AM5728 Cortex®-A15 Based Processors

Benefits

● Sitara’s highest performance ARM device with DSP accelerators

● 10,500 ARM Cortex-A15 DMIPS

● Enhanced graphics performance with HD video support

Sample Applications

● Human Machine Interface (HMI)

● Industrial PC

● Digital Signage

● High-end Navigation and Consumer

Key Features

● Dual 1.5GHz* Cortex-A15s, 213MHz M4s, and 750MHz** C66x DSPs

● 1080p Video Decode/Encode

● 3D (x2) and 2D acceleration

● Multiple Video Input Ports

● Display Subsystem

● Quad core Programmable Real-time Unit (PRU)/Industrial Communications Subsystem (2x dual core subsystems)

● USB3, PCIe, SATA

Power Estimates

● Typical Power: 2.5-4W (nominal voltage)

● Idle Power: ~650mW

Package: 23x23, 760 pin, 0.8mm

Availability:

● EVM: now

● Production: now