In the mobile world, voice is used for communicating as well as for interacting with mobile devices through Automatic Speech Recognition (ASR). Such devices must maintain a consistently high level of voice quality and intelligibility, regardless of the type and level of background noise. DSP Group®'s suite of HDClear-based patented audio/voice processing technologies dramatically reduces background noise during mobile conversations to deliver consistent voice clarity and to maximize ASR accuracy.
3D processing is used to enable true separation between voice and background noise. HDClear-based technologies further optimize the extracted voice for pure voice communications or as ASR input. As a result, background noise is dramatically reduced regardless of noise type, level and location (e.g. input angle), while a high and consistent degree of speaker identification and voice intelligibility is maintained.
Traditional background noise-cancellation techniques use two or more standard microphones in order to carry out beam forming or to mimic the human hearing system. The microphones capture similar information for voice and background noise, and therefore do not add significant “new information” for processing. As a result, these methods compromise on either voice quality or noise reduction. In contrast, HDClear-based technologies utilize an additional sensor, beyond the standard two audio microphones, for a third dimension of processing input. HDClear-based technologies gather environmental information during a call. A 3D-vocal algorithm is then applied to perform background noise reduction. In addition, multiple voice processing tasks are implemented, including Acoustic Echo Cancellation, loudness equalization, and more. The result is clearer, more intelligible voice communications on mobile devices, even in high ambient noise environments.
Noisy environments pose significant performance hurdles for ASR engines, which parse and translate voice input into text or control. By effectively filtering any type or level of ambient noise and delivering “cleaner” sound to the ASR engine, HDClear-based technologies markedly raise the ASR accuracy rate in any environment.
HDClear-based technologies enable the removal of background noise across all use cases, including a smartphone in handheld and speakerphone modes. This allows users to accurately carry out ASR tasks, or perform other operations using smartphone applications, during voice calls.
Background Noise Reduction in any Setting
Testing shows that HDClear-enabled devices experience only a minor degradation in very noisy environments and where SNR (voice-to-noise ratio) is low. In contrast, the performance of competitive noise-reduction solutions declines rapidly in noisy environments and where SNR is low, dropping as much as 20dB below that of HDClear-enabled devices. Since HDClear-based technologies perform consistently over a wide range of noise levels, users can depend on their smartphones in various noise settings.
Testing indicates that smartphones with HDClear-based technologies improve ASR accuracy compared to smartphones with their own built-in noise-reduction solution. As background noise increases, the accuracy rate of the HDClear-enabled device remains consistent, with only a minor degradation in noisy environments, while the second device degrades far more significantly in linear fashion. As a result, HDClear-enabled smartphone users can effectively carry out ASR functionality even in noisy environments.
Non-stationary and stationary noise reduction for both transmit and receive sides
Maximized Automatic Speech Recognition (ASR) accuracy rate
Acoustic Echo Cancellation (AEC)
Automatic Gain Control (AGC)
FlexiSpeech™ user-adjustable speech rate
Flexible Listening Enhancement (FLE)