Analysis of Time-Varying System Identification Using the Normalized Least Mean Square Algorithm in the Context of Data-Based Binaural Synthesis
Analysis of Time-Varying System Identification Using the Normalized Least Mean Square Algorithm in the Context of Data-Based Binaural Synthesis
复制标题
基于数据的双耳合成背景下使用归一化最小均方算法的时变系统辨识分析
DOI:
--
复制
发表时间:
2016
期刊:
影响因子:
--
通讯作者:
S. Spors
中科院分区:
文献类型:
--
作者:
Nara Hahn;S. Spors
The impulse response of a time-varying acoustic system can be measured by a continuous measurement technique. In such a measurement, a system is continuously excited by a periodic signal while the system changes over time, e.g. movement of source or receiver. The response of the system is captured by a microphone and the instantaneous impulse responses are computed from the microphone signal. Due to its time efficiency, continuous measurement methods are used for (i) the measurement of a large number of acoustic impulse responses e.g. spatial room impulse responses [1, 2] or head-related impulse responses [3]. During the measurement, the receiver (microphone or dummy head) moves on a predetermined trajectory. Depending on the required spatial resolution, an arbitrary number of impulse responses can be extracted and further used for sound field analysis or binaural synthesis [4, 5]. The second application of a continuous measurement technique is (ii) the auralization of dynamic auditory scenes [6]. For data-based binaural synthesis, for instance, binaural room impulse responses (BRIRs) are measured, and the ear signals are generated by filtering a dry source signal with the timevarying BRIRs. In this case, not only the accuracy of the individual impulse responses, but also the transient properties of the synthesized result have to be taken into account.