QNX® Aviage® Acoustic Processing Suite
The QNX Aviage Acoustic Processing Suite is a modular library of field-proven algorithms that improves the clarity, quality, and accuracy of voice communications for hands-free and speech recognition systems. Suited to diverse communication environments (including automotive, conference telephony, telepresence and VOIP), it allows customers to build a high-quality, resilient audio solution without expensive and time-consuming tuning for differing solutions within each environment.
Best-in-class audio quality
The QNX Aviage Acoustic Processing Suite contains modules on both the receive- and send-side that are designed specifically for clear and accurate sound delivery. They handle acoustic echo cancellation, noise reduction, wind blocking, dynamic parametric equalization, bandwidth extension, and more. Together these modules guarantee the finest possible quality communication — even in the challenging environment of a moving vehicle or crowded conference room.
Minimal tuning requirements
The QNX Aviage Acoustic Processing Suite is built to minimize the time and effort required for tuning, speeding products to market. Other features also streamline the tuning process:
- Comprehensive testing — testing against VDA (Verband der Automobilindustrie) standards, in-vehicle testing, robustness and regression, and an R&D phase multi-site test to ensure compliance to industry standards, and to reduce the tuning effort required for each project and implementation
- Tuning and integration tools — advanced diagnostics module and Windows remote-control client to analyze acoustic properties, speeding whatever final fine-tuning is required
- Scalable library of field-proven algorithms structured as pre-defined modules
- Simple API enabled by powerful and adaptive algorithms that minimize tuning
- Stand-alone C-callable library for shortest latency and lowest MIPS
- Advanced tools for more efficient integration and tuning
- Production proven technology — from Acura to Porsche
The QNX Aviage Acoustic Processing Suite offers lower production costs, easy customization and integration, and faster time to market.
Lower production costs
The QNX Acoustic Processing Suite eliminates the need for an additional, dedicated processor (DSP) for acoustic processing. This fact alone can generate savings of up to $15 per unit in Bill of Materials costs.
QNX Acoustic Processing Suite
The QNX Aviage Acoustic Processing Suite can run on the general purpose processor, saving the cost of the DSP.
The QNX Acoustic Processing Suite can offer these savings while delivering the high-quality, full-duplex hands free communication required today because it is a single library designed to co-exist with other applications on a general purpose processor or DSP. This library integrates easily and efficiently with the main CPU or DSP of a vehicle’s speech, telematics or radio module.
Easy customization and integration
Designed for use on multiple platforms and for easy upgrades, the QNX Aviage Acoustic Processing Suite can be reused as is, or easily customized across different product lines with minimal new development investment.
Tools delivered with the suite help developers and integrators complete their projects quickly and to the most exacting standards:
- Advanced diagnostics module — developers can quickly pinpoint and resolve problems
- Windows remote-control client — developers and acoustic engineers can find the optimal configurations by enabling them to turn features on or off, view and change parameters, such as EQ curves, read audio at every significant tap point, and more
Faster time to market
The QNX Aviage Acoustic Processing Suite speeds project development. Its modular architecture makes it easy to add, remove, customize and debug both receive-side and send-side processing steps.
Pre-testing ensures compliance to standards such as VDA (Verband der Automobilindustrie) and removes unknowns and variables that slow project development and drive up costs, while an Advanced Diagnostics module helps developers quickly identify and eliminate problems.
The QNX Aviage Acoustic Processing Suite is modular, so projects can implement all the functionality they need, but only the functionality they need.
The QNX Aviage Acoustic Processing Suite is designed and engineered to address the challenges of unforgiving acoustic environments. It provides the most sophisticated and flexible acoustic pre-processing solution available today.
QNX Aviage Acoustic Processing Suite Modules
QNX Aviage Acoustic Processing has many selectable modules for send and receive-side processing.
The QNX Aviage Acoustic Processing Suite is designed to enhance hands-free communication for the driver and the co-driver in an automobile, as well as for participants on the receive end of a conversation.
Modular software library
The sophistication of the QNX Aviage Acoustic Processing Suite comes from its high-quality, scalable software library of algorithms. This library is structured as pre-defined modules that can be configured based on application requirements, specific cabin conditions, and the level of voice enhancement required.
The library and modules can be initialized and applied to deliver excellent performance without any vehicle-specific tuning, or customized by developers to meet specific performance levels for designated environments.
The QNX Aviage Acoustic Processing Suite undergoes comprehensive testing, which ensures compliance to VDA, and upcoming JTU-T standards, and delivery of the best possible quality end-products, as well as reduced project risks.
VDA and in-vehicle testing
→ File-based VDA testing
→ In-vehicle VDA testing
→ Extensive tests in multiple vehicles prior to major releases
Robustness and regression
→ Extensive suite of “worst case scenario” tests for robustness
→ A suite of development, algorithmic, and QA regression tests
R&D phase multi-site tests
→ Diagnostic Rhyme Test (DRT)
→ Diagnostic Alliteration Tests (DALT) for intelligibility
→ CMOS quality tests
The QNX Aviage Acoustic Processing Suite send-side modules improve the audio signal captured by microphones in the vehicle in order to improve the quality of the signal received — and thus the audio — at the far-end.
Send-side modules used for acoustic processing
Acoustic echo cancellation
The acoustic echo cancellation (AEC) module reduces echo during hands-free calls by removing feedback generated by in-cabin speakers. It's algorithm is adapted to handle small and large cabin sizes, as well as automatically adjust for audio path delay.
Dual-channel mixer & post filter
The dual-channel mixer and post filter (CMIX) use information from two microphones to improve the speech components of a send signal, far more efficiently and effectively than multiple-microphone beam-forming solutions. The post filter provides spatial noise reduction by reducing uncorrelated noise picked up by the two microphones.
This module delivers optimal send-side speech quality for diverse microphone array designs. Tested with a wide range of array designs in different vehicles, with “worst-case” scenarios where microphones are shorted, this module offers the system designer greater flexibility in handling multiple customer microphone requirements than is possible with beam-former solutions.
Off axis rejection
The off axis rejection (OAR) module provides the ability to improve voice recognition performance in a multi-microphone environment. OAR module switches between the driver and co-driver depending on who is speaking and works to eliminate any other in-car noise or overlapping passenger speech. OAR is similar in concept to a beam-forming microphone although it is adaptable and uses less CPU horsepower.
High frequency encoding
The high frequency encoding (HFE) module compresses high frequency sounds into lower frequency sounds so that consonants (notably fricatives such as ‘s’ and ‘f’) transmit with improved intelligibility, but are still transmitted through cellular network codecs. This frequency encoding renders the telephone-transmitted voice both more natural and intelligible at the far-end.
Noise reduction & speech enhancement
The noise reduction (NR) and speech enhancement module reduces the impact of additive and dynamic noise, such as road noise. It preserves the natural quality, continuity, and flow of speech by removing unwanted noise and distortion, and enhancing the speech portion of the signal. The module includes dual microphone mixer support. New improvements to noise reduction provide for up to 20dB of noise removal without impacting speech.
Dynamic noise reduction
The dynamic noise reduction (DNR) module performs additional noise removal at lower frequencies. This helps minimize gating issues that can occur with CDMA phones. It also allows network voice codecs to properly detect pitch in noisy situations, without which voices can sound hoarse or unnatural. Additional features include tonal noise reduction, which detects and removes repetitive frequencies from the audio caused by grooved pavement, concrete separation strips, or engine noise.
Low frequency reconstruction
The low frequency reconstruction (LFR) module uses speech information present in the mid-frquency range to reconstruct speech frequencies masked by low frequency noise. Noise removal techniques remove the offending audio and LFR module adds back in the missing harmonics, resulting in more natural, fuller sounding speech.
Wind buffet removal
The wind buffet (WB) removal module identifies and eliminates the acoustical impact of wind flow from open windows, defrosters, drop-top convertibles, or building HVAC operations.
The equalization module controls the sound quality of the send portion of a hands-free conversation, using a parametric equalizer that supports both static (PEQ) and dynamic equalization (DPEQ). For DPEQ, the user can specify two static equalization curves; based on the signal-to-noise ratio at the microphone, the resulting equalization will be a smooth blend of both curves.
The dynamics module includes a smart automatic gain control (AGC) mechanism that increases or decreases gain of the signal, based on voice (not transient noises). It can do so independently for different sub bands so that voice is adjusted up and noise is adjusted down. The far end receives consistent volume level, increased perceived loudness and intelligibility, and reduced clipping. The soft limiter (LIM) prevents speech distortion caused by large peaks in speech.
The QNX Aviage Acoustic Processing Suite receive-side improve the quality of the received signal to offer constantly high-quality near-end audio, reducing the need for user interaction with system controls and presenting, not only a better communications experience, but also, in automotive environments, reducing driver distraction and improving safety.
Receive-side modules used for acoustic processing
Electrical noise suppresion (ENS)
The electrical noise suppresion module detects and suppresses noise between speech segments, such as from some NAD (network access device) chips.
Bandwidth extension (BWE)
The bandwidth extension module extends the narrow-band hands-free signal frequency range to deliver warm, natural and intelligible speechhelps deliver warm, bright and natural speech to the in-vehicle user.
It achieves a qualitative improvement in received speech quality thanks to a unique, patented technology that extends received signals beyond the narrow-band frequency range of 300 Hz to 3400 Hz used for hands-free telephone speech signals to the better approximate the full 60 Hz to 10 kHz range of natural human speech.
The equalization controls the sound quality of the receive portion of a hands-free conversation, using a parametric equalizer
Dynamics (AGC & LIM )
The dynamics module includes a smart automatic gain control (AGC) mechanism that increases or decreases gain of the signal, based on voice (not transient noises). The far end receives consistent volume level, increased perceived loudness and intelligibility, and reduced clipping. The soft limiter (LIM) prevents speech distortion caused by large peaks in speech.
Dynamic level control (DLC)
The dynamic level control module improves intelligibility in the presence of noise by allowing near end to hear speech at the same loudness across noise conditions.
Acoustic echo cancellation (AEC)
The acoustic echo cancellation module reduces echo during hands-free calls by removing feedback generated by in-cabin speakers.
Available with the QNX Aviage Acoustic Processing Suite are:
→ a Windows remote control client to facilitate system integration
→ an Advanced Diagnostics module to assist with system integration issues
The QNX Aviage Acoustic Processing Suite’s Windows remote control client is a Windows-based application for graphically controlling the behavior of the acoustic processing library. This module works over a TCP/IP connection with the hardware while a hands-free call is in progress, and does the following:
→ Permits turning on and off of features such as NR, AEC, RES, PEQ and AGC.
→ Views and changes parameters, such as EQ curves.
→ Records audio at various tap points to file.
→ Logs events as WAV file markers.
→ Saves and loads configurations.
→ Runs diagnostics with the Advanced Diagnostics module.
QNX Aviage Acoustic Processing Suite Windows Remote Control Client
Screen shot of the Windows remote control client used to control behavior of the acoustic processing library
The QNX Aviage Acoustic Processing Suite’s advanced diagnostics module supports automatic latency calibration, test waveform generation, coherence and distortion measurements, dynamic latency tests, amplifier dynamics checks, coupling measurements, and static and dynamic impulse response measurements.
For example, a file-based test application, which measures echo cancellation performance in the context of VDA specifications, isolates the echo cancellation software from the system as a whole to predict release performance ahead of production testing.
QNX Aviage Acoustic Processing Suite Advanced Diagnostics Module
Available as C-callable, stand-alone library.
32-bit general purpose fixed- and floating-point processors:
→ ARM 9/11
→ ARM Cortex A8/A9
→ PPC BE (big-endian)
→ PPC LE (little-endian)
→ SH2*, SH4
→ Jacinto (C64x)*
*supported as ports to other operating systems as specified by the customer
QNX SDP 6.4.0, or more recent.
Versions for other operating systems, such as Microsoft WinCE, Linux, or µITRON are available for selected architectures. Please contact your QNX sales person for more information.
Dependent on choice of sample rate and frame shift.
Hands-free mode requirements
→ 8 and 16 kHz, 16-bit PCM in/out
→ 128-point frame shift for 8 kHz
→ 256-point frame shift for 16 kHz
The QNX Aviage Acoustic Processing Suite is continuously evolving. Features provided in the last few releases include:
- Off axis rejection for speech recognition on dual-microphone systems
- Dynamic noise reduction and tonal noise reduction, and improved existing noise reduction algorithms
- Low frequency speech reconstruction
- Improvements to AGC algorithms to improve stability across various noise conditions
- New target hardware: SH2, x86, Cortex A8 fixed point
- Ability for diagnostic tool to listen to live audio tap streams
- Blended dual equalization curves (for both high- and low-noise environments) ensure removal of objectional spectral characteristics in high noise and a full audio spectrum in low noise
- Additional noise suppression to remove low-level electrical noise introduced by phone or network access device
- Increased handling of sampling frequencies (up to 16 kHz) to support new higher fidelity European mobile phone systems and better audio reproduction over MOST
- Improved performance during double-talk even in high coupling environments
- Enhancements to testing tools for easy test reproduction and in-vehicle harmonic distortion analysis
- Receive-side bandwidth extension (RBWE)
- Receive-side parametric equalization (RPEQ)
- Receive-side automatic gain control (RAGC)
- Receive-side dynamic limiter (RDL)
- Receive-side dynamic level control (RDLC)
- Additional diagnostics (DIAG)
- Improved source code examples
- Improvements to the Windows remote control client interface
- Dual-channel capability (MIX)
- High frequency encoding (HFE)
- Wind buffet suppression (WNDB)
- Send-side dynamic limiter (SDL)
- Diagnostics API (DIAG)
- Remote control server (RCS)
- Windows remote control client