Continuous rapid human voice source positioning method used for embedded platform

Abstract

The invention provides a continuous rapid human voice source positioning method used for an embedded platform. After ADC sampling is carried out by the embedded platform during phonation of a human, human voice digitalized threshold determining is carried out, and when a voice signal is greater than a threshold, algorithm operational data is acquired. The sampling of the various channels of the acquired is carried out, and calculation of time-delay between each two microphones is carried out. When the calculation of the time delay between each two microphones is completed, an asymptotic line equation of a hyperbola is used to estimate a direction angle. The direction angles are calculated according to the above mentioned time-delay calculation, and a final direction angle is estimated by adopting a centroid method. By adopting the continuous rapid human voice source positioning method, the number of the acquired voice signals is reduced under a condition of guaranteeing the positioning precision, and large-scale traversal calculation is prevented during the calculation of the time delay between each two microphones and the final direction angle, and therefore CPU resources are saved, and the real-time performance and the continuity of the voice source positioning angle acquisition are guaranteed.

Description

technical field [0001] The invention relates to a continuous and rapid human voice sound source localization method for an embedded platform. Background technique [0002] With the advancement of technology, sound source location has been more and more widely used in video conferencing, robot sound source location, bank security monitoring, sound-based tracking camera system, mechanical fault detection, noise source location, etc. However, due to the complexity of existing sound source localization algorithms, the current realization of sound source localization is based on PCs or high-performance dsp chips, which have high power consumption, high cost, many peripheral circuits, large volume, and high cost, resulting in some small volumes. However, it is difficult for large-scale application of sound source localization smart devices, especially small robots, with small battery capacity. [0003] In terms of calculating the time delay, the traditional sound source localizat...

AI Technical Summary

Problems solved by technology

However, due to the complexity of existing sound source localization algorithms, the current realization of sound source localization is based on PCs or high-performance dsp chips, which have high power consumption, high cost, many peripheral circuits, large volume, and high cost, resulting in some small volumes. , small battery capacity requires sound source localization smart devices, especially small robots, are difficult to apply on a large scale

[0003] In terms of calculating the time delay, the traditional sound source localization method usually uses two methods to calculate the maximum cross-correlation value of the two signals or to calculate the maximum point of the total energy value of the cross-power spectrum through traversal search when processing the time delay of two microphones. The delay value calculated by the former can only take the sampling time interval as the minimum unit, and the accuracy is poor; the latter is easily affected by noise and echo in practical applications, resulting in obvious calculation errors; and the calculation complexity of the two methods is high. Difficult to implement in a low-cost embedded environment

[0004] In terms of calculating the positioning angle through the time delay, the traditional sound source localization method usually uses three or more time delay differences and substitutes them into the hyperbolic equations for calculation, and the calculation complexity is high

[0005] In terms of microphone array design, the traditional sound source localization method mostly uses four or more pickups, and the rectangular layout or cross layout can easily obtain the sound delay and sound source localization angle between two microphones.

[0006] The above-mentioned design method directly leads to a large amount of calculation processing of the embedded platform, and the calculation time will be too long under the same main frequency embedded platform, and because the number of ADC samples is too large, the memory resources of the embedded platform are also relatively high. Generally, external memory is required

Due to the amount of calculation and resource consumption of existing sound source localization algorithms, existing methods can only be completed on some high-performance DSP processing chips on embedded platforms, accompanied by the addition of a large number of peripheral resources, so that some Equipment with sound source localization systems is expensive

Images