林麝应激反应声纹特征识别数据

应用场景说明：
1.自用场景：养殖场可通过声纹应激指数实时监测林麝应激状态，本数据集可为林麝环境系统的调节（如降噪或隔离应激源）提供支撑和依据，降低林麝应激发生率。
2.他用场景：①科研机构可利用高频能量占比与脉冲间隔的关联特征，为构建野生动物应激声学数据库提供数据支持。②音响设备商可依据典型应激声频谱特征（如4-8kHz频段），研发林麝专用白噪音发生器，降低突发噪音引发的应激反应。1.数据采集。数据采集自申请人自有林麝养殖基地，采集：日期、个体编号、声纹原始音频、发声持续时长（秒）、同期皮质醇浓度。剔除无效数据保证数据质量。
2.用Python librosa库进行声纹特征提取、分析：① 基频特征：提取音频基频均值（Hz），应激声＞800Hz；② 高频能量占比：计算4kHz-8kHz频段能量占总能量百分比；③ 声纹复杂度：通过梅尔频率倒谱系数（MFCC）标准差量化，值越高声纹变化越剧烈；④ 脉冲间隔：统计每500ms内声波脉冲峰值的平均间隔（ms）。
3.应激指数计算。应激指数=（基频/800 × 35）+（高频能量占比×25）+（声纹复杂度/100×20）+ （100/脉冲间隔×20）。（满分按100计）
4.应激等级判定。
① 正常：0＜应激指数≤40；
② 轻度应激：40＜应激指数≤70；
③ 高度应激：70＜应激指数≤100。
5.健康预警规则。单次高度应激且发声持续时长＞30秒或同期皮质醇浓度＞25ng/mL时触发健康预警。

Application Scenario Description:
1. Self-use Scenario: Forest musk deer breeding farms can monitor the stress status of forest musk deer in real time through the acoustic stress index. This dataset can provide support and basis for the adjustment of the forest musk deer environmental system (such as noise reduction or stressor isolation), reducing the incidence of stress in forest musk deer.
2. Other-use Scenarios: ① Research institutions can use the correlation features between high-frequency energy ratio and pulse interval to provide data support for the construction of wildlife stress acoustic databases. ② Audio equipment manufacturers can develop dedicated white noise generators for forest musk deer based on the typical stress sound spectrum characteristics (such as the 4-8kHz frequency band) to reduce stress reactions caused by sudden noise.

1. Data Collection. The data is collected from the applicant's own forest musk deer breeding base. The collected items include: collection date, individual ID, original voiceprint audio, sound duration (seconds), and simultaneous cortisol concentration. Invalid data is eliminated to ensure data quality.
2. Voiceprint feature extraction and analysis using the Python librosa library: ① Fundamental frequency feature: Extract the mean fundamental frequency of the audio (Hz); for stress sounds, the fundamental frequency is greater than 800Hz; ② High-frequency energy ratio: Calculate the percentage of energy in the 4kHz-8kHz frequency band relative to the total energy; ③ Voiceprint complexity: Quantified by the standard deviation of Mel-frequency cepstral coefficients (MFCC), where a higher value indicates more drastic changes in the voiceprint; ④ Pulse interval: Count the average interval (ms) of the peak values of acoustic wave pulses within every 500ms.
3. Stress Index Calculation. Stress Index = (Fundamental Frequency / 800 × 35) + (High-frequency Energy Ratio × 25) + (Voiceprint Complexity / 100 × 20) + (100 / Pulse Interval × 20). (Full score is 100)
4. Stress Level Judgment.
① Normal: 0 < Stress Index ≤ 40;
② Mild Stress: 40 < Stress Index ≤ 70;
③ Severe Stress: 70 < Stress Index ≤ 100;
5. Health Warning Rules. Trigger a health warning when a single severe stress event occurs with sound duration > 30 seconds or simultaneous cortisol concentration > 25ng/mL.

搜集汇总

背景与挑战

背景概述

该数据集聚焦于林麝的应激反应声纹特征识别，包含声纹原始音频、声纹特征参数（如基频均值、高频能量占比）以及基于算法计算的应激指数和等级，数据规模为724条并每日更新。它旨在通过声纹分析实时监测林麝的应激状态，支持养殖环境调节和科研数据库构建，具有明确的农业应用场景和算法规则。