Voice SDK

• Full-duplex echo cancelling, Echo Return Loss Enhancement (ERLE) > 45 dB, Category 1 of ITU-T P.340, compliant with VDA 1.6
• Single microphone noise reduction up to 20 dB (tune-able)
• Optional dual microphone array technology for acoustic focus on the user
• Support of 8kHz up to 24 kHz audio sampling rate
• Minimum tuning requirements–works out of the box
• Full parameter access for system tuning
• ARM Cortex NEON optimized code with small footprint

Acoustic Echo Cancelling (AEC) is one of the key components of Voice SDK. It is an extremely robust solution realizing full-duplex communication in a wide variety of use cases. It enables a VDA-compliant hands-free system with an Echo Return Loss Enhancement (ERLE) of at least 45 dB. Fully linear echo cancelling is combined with non-linear processing (residual echo suppression), maintaining full-duplex operation to the greatest possible extent.

Noise Reduction (NR) is the second core algorithm of an automotive hands-free SDK. With only one microphone, NR is capable of attenuating stationary or slowly modulating noise components up to 20 dB. This is achieved by a proprietary algorithm, which is capable of distinguishing between voice components and slowly modulating noise components, which are typical for automotive hands-free applications. The level of noise attenuation in dB is a tune-able parameter.

With two microphones, noise reduction can be further enhanced compared to a single microphone solution. The second microphone provides spatial information about situations of both voice signal and disturbances that come from different directions. As a result, non-stationary disturbances like interfering voices can be efficiently reduced. OpenSynergy offers different variants of array technologies for different microphone spacings and microphone types (conventional cardioid ECMs or modern MEMS microphones).

• Equalizers (receive (Rx) and transmit (Tx)) can compensate acoustic effects of vehicle interior
• Dynamic range compression increases the loudness of speech and avoids over modulation (clipping protection)
• Automatic Gain Control (AGC) adjusts the microphone gain for different loudness of user voices
• Noise dependent volume control increases loudspeaker output when the environmental noise level grows

A configuration and diagnostic link is very useful as tuning and monitoring interface, in order to check signal qualities in the digital domain, and to set parameter values (e.g. equalizer settings). OpenSynergy suggests a proprietary Bluetooth link to a PC for fast and convenient data exchange with the hands-free system. On the PC, a Graphical User Interface (GUI) is provided that helps to tune the system and – to a certain degree – can even perform an automatic assessment of hands-free operation quality.

• Removal of loudspeaker echoes (i.e. Acoustic Echo Cancelling = AEC) requires knowledge about the loudspeaker output source signal (= Echo Reference Signal)
• Echo Cancelling requires a synchronous audio loop, i.e. microphone(s) and loudspeaker A/D and D/A Codecs must operate on the same clock source
• Noise reduction (NR) with one microphone relies on algorithmic noise estimations. The clue is to distinguish between noise and voice components, which is difficult if noise is louder than voice (Signal-to-Noise-Ratio SNR <0 dB)
• An array of two microphones enables directional processing, e.g. voice of driver and passenger can be distinguished which is important for voice recognition

Voice SDK