Audio & media processors – Applications

DSPs for voice UI, control and audio and speech processing

TI has developed a flexible platform for voice processing on DSPs which work with multiple microphone configurations to help enable voice recognition in voice assistant applications. Field hardened algorithms for beamforming, noise reduction, dynamic range control and more are available in working project examples for evaluation and development. The software application is built on TI’s Processor SDK, allowing product choice and easy migration among TI’s processor portfolio.

Ultra-low power

At less than 150mW/MHz, the right fit for systems with two to four mics

Floating point signal processing

Mid-range DSP with performance for multi-microphone audio systems

High performance audio SoC

DSP-based SoC with MIPS and memory for higher complexity systems

Reference designs

TI’s audio pre-processing systems designs allow for easy evaluation and development of a DSP based application for far-field voice processing. Using field hardened, TI developed algorithms for beam-forming, noise reduction and other technologies, these designs show reference examples using different microphone array combinations to acquire the voice signal. The reference design also includes GUI based tools for calculating filter coefficients and easily configuring the input parameters to the DSP based on the microphone count and array geometry used. Learn more in the User’s Guides that come with the TI Designs.


Audio Preprocessing System Reference Design for Voice-Based Applications

This reference design uses multiple microphones and other processes to extract clear speech and audio amidst noise and other clutter.

Audio Pre-Processing System Reference Design for Voice-Based Applications

Voice-based design uses multiple microphones, a beamforming algorithm, and other processes to extract clear speech and audio amidst noise.

Audio Preprocessing System Reference Design for Voice-Based Applications

The TIDEP-0099 reference design uses a microphone array and a sophisticated signal processing to extract clear audio from noisy environments.

PCM1864-Based Linear Microphone Board (LMB) Reference Design

An easy-to-use reference design for applications that are cost sensitive and require clear-spoken audio for voice triggering and speech recognition.

PCM1864-Based Circular Microphone Board (CMB) Reference Design

A low-cost easy-to-use reference design for applications that require clear-spoken audio solutions.

Speech Recognition Reference Design on the C5535 eZdsp™

This reference design highlights the voice recognition capabilities of the C5535 and C5545 DSP devices using an embedded speech recognition library.


Processors ideal for pro audio and automotive audio amplifiers

 TI offers DSP and DSP+ARM® processors for a broad range of pro audio solutions from wireless microphones to multi-channel audio mixers as well as  optimized processor SOC architecture for automotive audio amplifiers.

Pro audio selection guide









Cores/speed C551x up to 200 MHz C55x up to 100 MHz 674x DSP at up to 456 MHz 674x DSP + ARM9 at up to 456 MHz 1 C66x DSP + ARM Cortex A15 at 600 MHz and 1 GHz 1 – 2 C66x cores at 600 MHz – 1.25 GHz 4 – 8 C66x cores at 1 GHz – 1.25 GHz 4 ARM Cortex A15 cores at 1.4 GHz, and 8 C66x DSP cores at 1.2 GHz
Performance Up to 400 MMACs 200 MMACs Up to 1912 FLOPs, 3468 MACs Up to 1912 FLOPs, 3468 MACs Up to 2100 DMIPS, 19.2 GMAC, 28.8 GFLOPS Up to 80 GMACs, 40 GFLOPS Up to 320 GMACS, 160 GFLOPS Up to 19,600 DMIPs, 307.2 GMACs, 198.4 GFLOPs
Key peripherals McBSP USB McASP, 10/100 Enet, USB McASP, 10/100 Enet, USB McASP, McBSP, USB, GEnet SGMII, 2 x PCIe 2 x SGMII, 2 x PCIe, 4 x SRIO 4 x GbEnet, 2 x 10 GbEnet, SRIO
Power .54W/MHz active .33mW/MHz active <.5W <1W 2W – 4W 1.5W – 3W 6W – 10W 10W – 15W
Extended temp range -40 – 100C - 40 – 85C -40 – 105C -40 – 105C -40 – 125C -40 – 100C -40 – 100C -40 – 100C
Package 10mm2 12 mm2 13mm2 and 16mm2 13mm2 and 16mm2 21mm2 21mm2 24mm2 40mm2
Price From: $4.17 @1ku From: $1.95 @ 1ku From: $6.30 @ 1ku From: $9.58 @ 1ku From: $15.39 @ 1ku From: $24.90 @1ku From: $125.00 @ 1ku From: $661.26 @ 1ku

Pro audio benchmark starter kit

The audio benchmark starter kit provides a quick and easy way to benchmark key audio functions on TI’s DSP devices. This starter kit is intended to help all users, especially those new to the TI DSP development environment, with the following:

  • Become familiar with benchmarking functions on TI DSPs.
  • Understand compiler options and code/data memory placement that allows developers to obtain optimal performance on TI’s DSP Architectures.
  • Utilize signal processing functions from TI’s signal processing libraries, which contains several kernels optimized for TI DSP architectures.
  • Read the full application note or view the training video.

The following table shows the functions benchmarked along with the results obtained:

Processor core
C66x DSP Core
Hardware platform used 66AK2G12 EVM
Devices featuring benchmarked core C66x DSPs
66AK2x processors
Sitara AM57x processors
Function benchmarked C66x execution time Associated TI library
C66x cycles C66x μS @ 600 MHz
Single Precision FFT (256 samples) 1808 3.01 DSPLIB
Single Precision FFT (1K samples) 8663 14.44 DSPLIB
Single Precision FIR (128 samples, 16 coefficients) 2652 4.42 DSPLIB
Single Precision IIR
(1k samples from 2 channel with 3 stage cascade biquad)
8258 13.76 DSPLIB

Audio benchmarks starter kit

The audio benchmarks starter kit is part of the processor SDK RTOS release.

66AK2Gx software development kit

Software development kit (SDK) for 66AK2Gx processors - Linux and TI-RTOS support.

AM57x Sitara™ software development kit

Processor SDK is a unified software platform for TI embedded processors providing easy setup and fast out-of-the-box access to benchmarks and demos.

View the applications available for Audio & media processors in these areas: