MediaTek AIoT Smart Speaker

Meeting the demands of both cost-effective and high performance far-field voice products, the MediaTek Pumpkin i300B Reference design delivers a rich set of audio features powered by MediaTek 8632B SoC. The integrated TalkTo Voice UI passes both the AVS Far-field 1.7 and Google Voice Assistant (GVA) test requirements, accelerating the development of high performance smart speaker and home audio products.


  • Reliable Far-field performance
    • DSP Concepts’ cutting-edge algorithms to cut through the loud music or kitchen noises in living room
    • AVS 1.7 far-field and Google ART certified algorithms
    • Best-in-class barge-in and wake word detection performance
  • Support for high resolution audio playback
    • Multi-channel audio playback for immersive sound applications on smart speakers and sound bars
    • 3 channel AEC to effectively cancel out the multi-channel playback signals

Audio Front-end Features

  • DSP Concepts’ TalkTo for excellent far-field performance in noisy environments
    • 4-mic Beamformer with Single Channel Noise Reduction (SCNR)
    • State-of-the-art noise cancellation with Adaptive Interference Canceller (AIC) 
      • No reference signal needed
      • Eliminates 25+ dB of external noise sources such as background music or a booming movie
    • 30+ dB Acoustic Echo Cancellation (AEC) for excellent music barge-in
  • Support for 3.1 channel playback processing 

Software Integrations

  • Audio Weaver Designer: Drag-and-drop graphical tool to customize and tune audio processing models in real-time with 500+ modules
  • Audio Weaver Core: Embedded audio processing engine for Cortex-A35
  • TalkTo Voice UI: Noise reduction, echo cancellation and advanced beam-forming signal processing for far-field audio detection
  • Multi-channel playback algorithms

Hardware Specifications

  • Processor: MT8362B
    • Quad-core 64-bit ARM-Cortex A35 MPCoreTM at 1.3 GHZ 
  • Memory
    • 32-bt LPDDR3/DDR3L up to 800MHz
    • 32-bit DDR up to 667MHz
    • e.MMC; SLC NAND Flash
  • Audio I/O
    • Up to 3 analog or 2 digital microphones, 8-ch TDM for microphone array, I2S x2, SPDIF audio inputs
    • Line out, Headphone, 8-ch TDM, I2S x2 and S/PDIF audio outputs
    • 4-ch I2Sx2, 8-ch TDM and 2-ch PDM audio inputs for microphone voice input control and connected audio products
  • Connectivity
    • WiFi-4 (b/gn) 2.4GHz and Bluetooth 4 (Low Energy) connectivity
    • WiFi-5 (b/g/n/ac) Dual-band 2x2 & Bluetooth 5 via MT7668

Included with the demo kit:

  • MediaTek AIoT Pumpkin Evaluation Kit with i300B SoC
  • Vesper HAT with 4-mic Trillium microphone array geometry integrated with dust and water resistant piezoelectric MEMS microphones
  • Voice UI Quickstart Bundle: Everything you need to know to design voice-enabled Smart home and IoT products, including design guidelines,microphone placement, an Audio Weaver Designer layout (.AWD), and tools to get started.
  • USB Cables

Ready to get started?

Schedule time to speak with one of our audio experts.