Digital Signal Processor (DSP)
Fortemedia’s core competency is voice processing solutions for consumer applications. Our solutions provide voice enhancement in the areas of Far Field Pickup, improved Voice Recognition performance as well as unique low power solution for hands free device control and voice wake-up. Fortemedia’s voice processing solutions are delivered either as embedded software on popular codecs and Application Processor or as very cost-effective and easy-to-integrate IC packages.
iM401
Smart Microphone Processor
Introduction: The iM401 is a mini DSP (digital signal processor) designed for voice trigger processing. The iM401 is an all-digital product that supports digital microphone input. The iM401 is built to support not only voice trigger and key phrase detection but also one-shot command recognition with 3rd party algorithms.
Summary of Benefits:- Low power operation during silence and voice activity detection
- Option for very low power digital voice activity detection
- Selectable thresholds for false detection and false recognition rate balancing
- OEM defined trigger words and phrases supported
- Hands free wake-up of battery powered devices with extremely low power
- Tensilica HiFi series DSP Core
- Programmable operation speed for variety of power saving modes
- Dedicated memory built in for voice buffer, can also be used as runtime program memory
- Power management and enhanced power saving
- Stereo digital microphone inputs
- Interrupt (IRQ) output to indicate a voice wake up event to host chip
- Single digital power supply from 1.6V to 3.6V voltage
- Pre-trained trigger word support – multiple languages
- Option for user defined trigger word or phrase
- 2 to 3 seconds Voice Buffer capability
- I2C Control Interface with simple IRQ interface to wake up Application Processor
iM501
Smart Microphone Processor
Introduction: The iM501 integrates a Tensilica HiFi series DSP core to provide voice-triggered wakeup function and post-wakeup voice processing. Working in conjunction with the Fortemedia microphone processor, iM205, iM501 can provide an ultimate power saving solution. There are multiple voice-trigger engines, smart microphone controls, and a minimum of 3 seconds of voice buffer that could co-exist in the system. Together with the Fortemedia advanced microphone voice-enhancement technology, iM501 is a versatile Voice Interface Device improving both human-to-machine and human-to-human voice communications.
Summary of Benefits:- VOS (Voice Operating System) for always-on ambience monitoring
- Mechanism of ultra-low current for wakeup function
- Capability of beamforming, noise suppression and acoustic echo cancellation
- Voice Buffer Scheme
- Provide more than 3 seconds of dual voice buffer for seamless operation
- Hardware assisted auto-buffers for MICs, DAC/TXs and RXs data to reduce system overhead
- Multiple microphone output formats and interfaces
- Pre-trained trigger word support – multiple languages
- Option for user defined trigger word or phrase
- 3 seconds of Voice Buffer capability
- I2C Control Interface with simple IRQ interface to wake up
- Application Processor
iM505
Smart Microphone Processor
Introduction: The iM505, Fortemedia’s latest Voice Interface Device (VID), is an ultra-low power voice processor featuring voice fundamentals with “Always-on Acoustic Awareness, wake-up and keyword recognition along with versatile voice processing. The solution provides Voice Operation System (VOS) that pushes the system performance to ultra-low power consumption while asleep with its custom core design and accelerates times to market with its unique combined hardware and software package. The iM505 consists of Tensilica DSP with computation up to 300+MIPS and an optimized AI Engine that is best-suited to neural network voice application.
Summary of Benefits:
- Time-to-market: VOS infrastructure & Test Bench/Virtual Lab
- VOS can be used with the iM505 to provide users with a thin-layered and efficient real-time environment that integrates specific algorithms for a variety of applications
- System standby at ultra-low power mode
- Less than 10uA leakage current when iM505 is sleeping. When combined with sam (small array microphone)’s AVD (Analog Voice Detection) feature, it consumes ultra-low power in a quiet environment
Features:
- Tensilica DSP inside with programmable MIPS: Low MIPS (<50MIPS) and High MIPS (up to 300 MIPS)
- AVD trigger wake up to provide ultra-low current for wakeup function in smart mic
- Voice Buffer Scheme
- Ultra-low Latency Time ~5ms from triggering to DSP ready
- Equip 512KB ROM, enabling Fortemedia’s VIP Tower foundations: AoAA, VOS, AIE and sam
- Built-in AI engine with 32 sets of MACs for parallel computing
- Multiple sets of audio interface include up to 8 channels PDM input, 3 sets of TDM, I2S, and dual channels PDM output
- VOS (Voice Operating System) to facilitate applications into iM505CE-M
- Voice stream infrastructure with HAL built-in library for application portability
- Capability of inter-chip communication among multiple iM505, smart mic, and VIDs
- Programmable speech detection and voice triggering
- Comprehensive voice processing, covering beamforming, noise suppression and acoustic echo cancellation