Sound + Technology

  • 01

    Echo Cancellation(AEC)

    Echo Cancellation works on either mono                or stereo playback

  • 02

    Dereverberation

    Remove the confusing trailing residual                in voice recording which is caused by                room reflections

  • 03

    Sound Source Localization

    Accurately locate the target voice direction

  • 04

    Beam Forming

    Greatly improve the accuracy of far-field speech recognition

  • 05

    Noise Suppression

    Remove steady noise such as those from  air-conditioner or vehicle

  • 06

    Keyword spotting(KWS)

    Rapid responding and Supports multi-layer voice interaction

CORE TECHNOLOGIES

  • Echo Cancellation AEC

    Echo Cancellation works on either mono or stereo playback

    支持单声道和立体声回声抵消。支持大音量、重低音情况下的回声抵消,可实现音响大音量(100dB)情况下非线性噪声残留抑制,可以解决在麦克风与大音量喇叭距离较近时(比如一些小型智能音箱)以及一些低成本喇叭失真所带来的AEC性能下降的痛点。

  • Dereverberation

    Remove the confusing trailing residual in voice recording which is caused by room reflections

    房间混响是在封闭空间内进行语音通信时,语音信号从声源到达传声器的多径反射所造成的,其中到达传声器较晚的反射波称作晚期混响,晚期混响会对语音造成自掩蔽和重叠掩蔽效应,因而影响语音的可懂度。语音去混响技术可以去除房间混响导致的语音拖尾。

  • Source Localization

    provide all omni-direction, accurate localization of sound sources

    利用麦克风阵列,实现360°全方位语音信号采集,并通过声源定位来确定目标说话人的方向,支持多声源情况的定位,根据语音及唤醒词特性,可在干扰噪声存在时准确定位目标语音方向。

  • Beam Forming

    Greatly improve the accuracy of far-field speech recognition

    在保护主方向语音的同时,去除环境中的语音、电视噪声等非平稳干扰,其中非平稳噪声抑制量大于20dB,在噪音环境下可以大幅提升远场语音识别率。

  • Noise Suppression

    Remove steady noise such as those from air-conditioner or vehicle.

    去除空调、胎噪等平稳噪声以及环境中的语音、电视等非平稳噪声的波束成形算法。

  • Keyword spotting KWS

    Rapid responding and Supports multi-layer voice interaction

    用户通过说出关键词,可以将麦克风阵列从休眠状态唤醒。目前支持Dual-wake、Free-cut、One-shot等独有的定制功能,具有低于500毫秒的机器应答响应的极致速度,同时支持多轮对话功能。