'[2000;1000,1000;500,500,1000;2000]',

'[1000,1000;500,500,1000;1000,1000;2000]',

'[1000,--(1000);1000,--(1000);500,250,250,1000;--(1000),1000]',

'[500;1000,500,1000,500;500,500,500,250,250,500,500;250,250,500,500,1000]',

'[1000,--(1000);1000,--(1000);1000,-(500),500;1000,1000]',

'[750,250,500,500,500,-(500);500,1000,500,500,-(500);750,250,500,500,500,-(500)]',

'[500,1000,500,500,250,250;1000,500,750,250,500;3000]',

'[500,500,500;1000,500;500,500,500;1500;500,500,500;1000,500;500;1000;1500]',

'[500,500,1000;500,500;1000;375,125,250,250,375,125,250,250;500,500,1000]',

'[500,500,1000;500,500,1000;500,500,750,250;2000]',

'[1000,1000;500,500,1000;1000,500,500;2000]',

'[1000,1000;500,500,1000;500,250,250,250;2000]',

audio, sr = librosa.load(path)

energy = librosa.feature.rmse(audio)

frames = np.nonzero(energy >= np.max(energy) / 5)

indices = librosa.core.frames_to_samples(frames)[1]

audio = audio[indices[0]:indices[-1]] if indices.size else audio[0:0]

c_max = np.argmax(chromagram, axis=0)

#print(c_max.shape[0])

#print(c_max)

# print(np.diff(c_max))

# chromagram_diff = np.diff(chromagram,axis=0)

# print(chromagram_diff)

# sum_chromagram_diff = chromagram_diff.sum(axis=0)

# test = np.array(sum_chromagram_diff)

# plt.plot(test)

img = np.zeros(chromagram.shape, dtype=np.float32)

w, h = chromagram.shape

for x in range(h):

# img.item(x, c_max[x], 0)

img.itemset((c_max[x], x), 1)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

raw_rms = rms.copy()

rms = np.diff(rms)

rms_on_onset_frames_cqt = [rms[x] for x in onset_frames_cqt if x < len(rms)]

min_rms_on_onset_frames_cqt = np.min(rms_on_onset_frames_cqt)

rms = [1 if x >= min_rms_on_onset_frames_cqt else 0 for x in rms]

for i in range(1,len(rms)):

tmp = [np.abs(i - x) for x in onset_frames_cqt]

min_gap = np.min(tmp)

sub_onset_frames_cqt = [x for x in onset_frames_cqt if x<i]

if len(sub_onset_frames_cqt) > 0:

last = sub_onset_frames_cqt[-1]

start = last

end = i

threshold_rms = threshold * np.max(raw_rms)

if raw_rms[start]<threshold_rms and raw_rms[start+1]>threshold_rms:

start += 1

if start <= end:

continue

sub_rms = raw_rms[start:end]

min_rms = np.min(sub_rms)

if rms[i] == 1 and rms[i-1] == 0 and min_gap > 5 and raw_rms[i] > threshold_rms and min_rms< threshold_rms:

#print("start,end,min_rms,threshold_rms is {},{},{},{}".format(start,end,min_rms,threshold_rms))

is_onset = check_onset_by_cqt(y, onset_frames_cqt, i)

if is_onset:

onset_frames_cqt.append(i+1)

else:

if rms[i] == 1 and rms[i-1] == 0 and min_gap > 5 and raw_rms[i] > np.max(raw_rms) * threshold:

is_onset = check_onset_by_cqt(y, onset_frames_cqt, i)

if is_onset:

onset_frames_cqt.append(i + 1)

onset_frames_cqt.sort()

if len(onset_frames_cqt) < 0:

return onset_frames_cqt

cqt = librosa.amplitude_to_db(librosa.cqt(y, sr=16000), ref=np.max)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

cqt[0:30, :] = -100

w,h = cqt.shape

max_cqt = [np.max(cqt[:, x+2]) for x in onset_frames_cqt if x < h-2]

mean_max_cqt = np.mean(max_cqt)

global_before_cqt = [np.max(cqt[:, x-6:x - 3]) for x in onset_frames_cqt if x > 6]

mean_global_before_cqt = np.mean(global_before_cqt)

#print("mean_max_cqt is {}".format(mean_max_cqt))

step = 4

w,h = cqt.shape

result = []

for i in range(step,h-4,3):

before_cqt = [np.max(cqt[:, x]) for x in range(i-step,i)]

#before_cqt = [cqt[:, x] for x in range(i - step, i)]

mean_before_cqt = np.mean(before_cqt)

after_cqt = [np.max(cqt[:, x]) for x in range(i, i + step)]

#after_cqt = [cqt[:, x] for x in range(i, i + step)]

mean_after_cqt = np.mean(after_cqt)

#if np.abs(mean_after_cqt - mean_max_cqt) < 5:

if np.abs(mean_after_cqt - mean_max_cqt) < 10 and mean_before_cqt < mean_max_cqt and mean_before_cqt < mean_global_before_cqt + 7 and np.abs(mean_before_cqt - mean_global_before_cqt) < 10 and rms[i] < rms[i+1] and rms[i] > np.max(rms)*0.2:

#print("mean_before_cqt,mean_global_before_cqt,mean_after_cqt,mean_max_cqt,i is {},{},{},{},{}".format(mean_before_cqt,mean_global_before_cqt,mean_after_cqt,mean_max_cqt,i))

# cqt上半部存在亮的水平线

#if i + 2 < h and ( np.abs(np.max(cqt[30:, i + 2]) - mean_max_cqt) < 10 or np.max(cqt[30:, i + 2]) > mean_max_cqt):

result.append(i)

if result:

min_width = 5

# print("min_width is {}".format(min_width))

result = del_overcrowding(result, min_width)

for i in result:

tmp = [np.abs(i - x) for x in onset_frames_cqt]

min_gap = np.min(tmp)

if min_gap > 5:

onset_frames_cqt.append(i)

onset_frames_cqt.sort()

cqt = librosa.amplitude_to_db(librosa.cqt(y, sr=16000), ref=np.max)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

cqt[0:30, :] = -100

w, h = cqt.shape

max_cqt = [np.max(cqt[:, x + 2]) for x in onset_frames_cqt if x < h - 2]

mean_max_cqt = np.mean(max_cqt)

result = []

for x in onset_frames_cqt:

# cqt上半部存在亮的水平线

print("real,mean_max_cqt,end is {},{},{}".format(np.max(cqt[30:, x + 2]), mean_max_cqt, x))

#if x + 2 < h and (np.abs(np.max(cqt[30:, x + 2]) - mean_max_cqt) < 5 or np.max(cqt[30:, x + 2]) > mean_max_cqt):

result.append(x)

cqt = librosa.amplitude_to_db(librosa.cqt(y, sr=16000), ref=np.max)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

cqt[0:30, :] = -100

max_cqt = [np.max(cqt[:, x+2]) for x in onset_frames_cqt]

mean_max_cqt = np.mean(max_cqt)

#print("mean_max_cqt is {}".format(mean_max_cqt))

global_before_cqt = [np.max(cqt[:, x-5:x - 2]) for x in onset_frames_cqt]

mean_global_before_cqt = np.mean(global_before_cqt)

#print("mean_global_before_cqt is {}".format(mean_global_before_cqt))

step = 4

w,h = cqt.shape

result = False

i = onset_frame

before_cqt = [np.max(cqt[:, x]) for x in range(i-step,i)]

#before_cqt = [cqt[:, x] for x in range(i - step, i)]

mean_before_cqt = np.mean(before_cqt)

after_cqt = [np.max(cqt[:, x]) for x in range(i, i + step)]

#after_cqt = [cqt[:, x] for x in range(i, i + step)]

mean_after_cqt = np.mean(after_cqt)

#if np.abs(mean_after_cqt - mean_max_cqt) < 5:

if np.abs(mean_after_cqt - mean_max_cqt) < 10 and mean_before_cqt < mean_max_cqt and np.abs(mean_before_cqt - mean_global_before_cqt) < 5 and rms[i] < rms[i+1] and rms[i] > np.max(rms)*0.2:

#print("mean_before_cqt,mean_global_before_cqt,mean_after_cqt,mean_max_cqt,i is {},{},{},{}".format(mean_before_cqt,mean_global_before_cqt,mean_after_cqt,mean_max_cqt,i))

result = True

cqt = librosa.amplitude_to_db(librosa.cqt(y, sr=16000), ref=np.max)

cqt[0:30, :] = -100

w,h =cqt.shape

max_cqt = [np.max(cqt[:, x + 2]) for x in onset_frames_cqt if x < h-2]

mean_max_cqt = np.mean(max_cqt)

global_max = np.max(cqt)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

min_waterline = find_min_waterline(rms, 8)

gap = 0

if len(min_waterline) > 0:

waterline = min_waterline[0][1]

gap = (np.max(rms) - waterline) * 0.2

# 关于第一个节拍

# 条件一：节拍点的前后有高差，前小后大

#print("checking the first")

condation1 = rms[onset_frames_cqt[0] + 1] - rms[onset_frames_cqt[0] - 1] > 0.12 * np.max(rms)

#print("condation1 is {},{},{}".format(condation1,rms[onset_frames_cqt[0] + 1],rms[onset_frames_cqt[0] - 1]))

# 条件三：后面的cqt上半部存在亮的水平线

#condation3 = np.abs(np.max(cqt[30:, onset_frames_cqt[0] + 2:onset_frames_cqt[0] - 6]) - global_max) < 10

condation3 = True

#print("condation3 is {}".format(condation3))

# 条件四：前面的cqt上半部与后面的cqt上半部有差别

condation4 = np.abs(np.mean(cqt[30:, onset_frames_cqt[0] - 6:onset_frames_cqt[0] - 2]) - np.mean(cqt[30:, onset_frames_cqt[0] + 2:onset_frames_cqt[0] + 6])) > 4

#condation4 = True

tmp1 = onset_frames_cqt[0] - 10 if onset_frames_cqt[0] - 10 >= 0 else 0

tmp2 = onset_frames_cqt[0] - 2 if onset_frames_cqt[0] - 2 >= 0 else 0

tmp3 = onset_frames_cqt[0] + 2 if onset_frames_cqt[0] + 2 <= len(rms) else len(rms)

tmp4 = onset_frames_cqt[0] + 8 if onset_frames_cqt[0] + 8 <= len(rms) else len(rms)

if tmp1 >= tmp2:

condation4 = False

else:

condation4 = np.abs(np.mean(cqt[30:, tmp1:tmp2]) - np.mean(cqt[30:, tmp3:tmp4])) > 3

#print("condation4 is {}".format(condation4))

if condation1 and condation3 and condation4:

result = [onset_frames_cqt[0]]

else:

result = []

for i in range(1,len(onset_frames_cqt)):

start = onset_frames_cqt[i-1]

end = onset_frames_cqt[i]

sub_rms = rms[start:end]

#print("checking {}".format(end))

# 条件一：节拍点的前后有高差，前小后大

#print("len(rms),end is {},{}".format(len(rms),end))

tmp1 = len(rms)-1 if end >=len(rms)-1 else end - 1

tmp2 = len(rms)-1 if end >=len(rms)-2 else end + 1

condation1 = rms[tmp2] - rms[tmp1] > 0.12 * np.max(rms)

#print("condation1 is {}".format(condation1))

# 条件二：跟前一节拍之间有波谷

#condation2 = np.min(sub_rms) < np.max(rms) * 0.4

condation2 = True

#print("condation2 is {}".format(condation2))

#条件三：后面的cqt上半部存在亮的水平线

tmp1 = len(rms)-1 if end + 2 >= len(rms) else end + 2

tmp2 = len(rms)-1 if end + 6 >= len(rms) else end + 6

if tmp1 >= tmp2:

condation3 = False

else:

condation3 = np.abs(np.max(cqt[30:, tmp1:tmp2]) - global_max) < 10

# print("condation3,np.max(cqt[30:, end + 2:end + 6]) - global_max is {},{},{}".format(np.max(cqt[30:, tmp1:tmp2]),global_max,condation3))

# 条件四：前面的cqt上半部与后面的cqt上半部有差别

tmp1 = len(rms)-1 if end + 2 >= len(rms) else end + 2

tmp2 = len(rms)-1 if end + 6 >= len(rms) else end + 6

if tmp1 >= tmp2:

condation4 = False

else:

condation4 = np.abs(np.mean(cqt[30:, end - 6:end - 2]) - np.mean(cqt[30:, tmp1:tmp2])) > 3

#print("condation4 is {}".format(condation4))

#if end < len(rms) -6 and np.min(sub_rms) < waterline + gap and (rms[end + 1]>threshold * np.max(rms) or rms[end + 2]>threshold * np.max(rms)) and rms[end + 1] > np.max(rms) * 0.15 :

if end < len(rms) - 6 and condation1 and condation2 and condation4:

# cqt上半部存在亮的水平线

#print("real,mean_max_cqt,end is {},{},{}".format(np.max(cqt[30:, end + 2]),mean_max_cqt,end))

#if np.abs(np.max(cqt[30:, end + 2]) - mean_max_cqt) < 4:

# is_onset = check_onset_by_cqt(y, onset_frames_cqt, end)

# if is_onset:

result.append(end)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

result = []

for x in onset_frames_cqt:

# 较大上升沿 或 波谷点

if rms[x+1] - rms[x] > threshold1 or (rms[x-1] - rms[x] > threshold2 and rms[x+1] - rms[x] > threshold2):

result.append(x)

onsets_frames = get_real_onsets_frames_rhythm(y, modify_by_energy=True, gap=0.1)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

rms_on_onsets = [rms[x] for x in onsets_frames]

mean_rms = np.mean(rms_on_onsets)

best_threshod = mean_rms/np.max(rms)*0.4

#best_threshod = np.min(rms_on_onsets)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

base_line = np.max(rms) * best_threshod

onsets_on_best_threshod = [i for i in range(1,len(rms)-1) if rms[i] <= base_line and rms[i+1] >= base_line]

for x in onsets_on_best_threshod:

offset = [np.abs(x - i) for i in onsets_frames]

min_gap = np.min(offset)

if min_gap > 5:

onsets_frames.append(x + onsets_frames[0])

onsets_frames.sort()

y, sr = librosa.load(filename)

#type_index = get_onsets_index_by_filename(filename)

total_frames_number = get_total_frames_number(filename)

# base_frames = onsets_base_frames_rhythm(type_index,total_frames_number)

base_frames = onsets_base_frames(onset_code, total_frames_number)

# 标准节拍个数

topN = len(base_frames)

gap = 0.1

run_total = 0

onset_frames_cqt = []

#threshold = 0.35

best_onset_frames_cqt = []

best_total = 0

threshold = get_best_threshod(y)

#print("best threshold is {}".format(threshold))

onset_frames_cqt = get_real_onsets_frames_rhythm(y, modify_by_energy=True, gap=gap)

#waterline, best_starts_waterline = find_best_waterline(rms, 4, topN)

while True:

# 从CQT特征上获取节拍

#onset_frames_cqt = get_real_onsets_frames_rhythm(y, modify_by_energy=True,gap=gap)

if onset_frames_cqt:

min_width = 5

#print("min_width is {}".format(min_width))

onset_frames_cqt = del_overcrowding(onset_frames_cqt, min_width)

#print("0. onset_frames_cqt is {}".format(onset_frames_cqt))

# CQT = librosa.amplitude_to_db(librosa.cqt(y, sr=16000), ref=np.max)

# onset_frames_cqt = get_miss_onsets_by_cqt(CQT, onset_frames_cqt)

# if len(onset_frames_cqt)<1:

# onset_frames_cqt = get_real_onsets_frames_rhythm(y, modify_by_energy=True)

if len(onset_frames_cqt) > 0:

min_width = 5

#print("min_width is {}".format(min_width))

onset_frames_cqt = del_overcrowding(onset_frames_cqt, min_width)

#print("1. onset_frames_cqt is {}".format(onset_frames_cqt))

# 根据rms阀值线找漏的

if len(onset_frames_cqt) > 0:

onset_frames_cqt = get_missing_by_best_threshod(y, onset_frames_cqt, threshold)

# 去伪

if len(onset_frames_cqt) > 0:

onset_frames_cqt = find_false_onsets_rms(y, onset_frames_cqt, threshold)

#print("2. onset_frames_cqt is {}".format(onset_frames_cqt))

# 找漏的

if np.abs(len(onset_frames_cqt) - topN) <=3 and len(onset_frames_cqt) > 0:

onset_frames_cqt = get_miss_onsets_rms(y, onset_frames_cqt,threshold)

onset_frames_cqt = get_miss_onsets_by_cqt(y, onset_frames_cqt)

#print("3. onset_frames_cqt is {}".format(onset_frames_cqt))

if len(onset_frames_cqt) >= best_total:

best_total = len(onset_frames_cqt)

best_onset_frames_cqt = onset_frames_cqt

#print("4. onset_frames_cqt is {}".format(best_onset_frames_cqt))

if len(onset_frames_cqt) - topN >= 0 and len(onset_frames_cqt) - topN <= 3 or run_total >0:

#print("best_onset_frames_cqt,len, run_total is {},{},{}".format(best_onset_frames_cqt,len(onset_frames_cqt),run_total))

break

else:

threshold *= 0.9

run_total += 1

#type_index = get_onsets_index_by_filename(filename)

total_frames_number = get_total_frames_number(filename)

# base_frames = onsets_base_frames_rhythm(type_index,total_frames_number)

base_frames = onsets_base_frames(onset_code, total_frames_number)

# 标准节拍个数

topN = len(base_frames)

y, sr = librosa.load(filename)

onsets_frames = get_real_onsets_frames_rhythm(y, modify_by_energy=True, gap=0.1)

if onsets_frames:

min_width = 5

# print("min_width is {}".format(min_width))

onsets_frames = del_overcrowding(onsets_frames, min_width)

#print("0. onset_frames_cqt is {}".format(onsets_frames))

# 如果已经匹配很好，就直接返回

if len(onsets_frames)>0 and len(onsets_frames) == topN:

base_frames = onsets_base_frames(onset_code, total_frames_number - onsets_frames[0])

base_frames = [x + (onsets_frames[0] - base_frames[0]) for x in base_frames]

min_d, best_y, modify_onsets_frames = get_dtw_min(onsets_frames.copy(), base_frames, 65)

#print("min_d is {}".format(min_d))

if min_d < 4:

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

rms_on_onsets = [rms[x] for x in onsets_frames]

mean_rms = np.mean(rms_on_onsets)

threshold = mean_rms / np.max(rms)

#threshold = np.min(rms_on_onsets)

return onsets_frames,threshold

best_onset_frames_cqt = []

best_total = 0

best_threshold = 0

onset_frames_cqt,threshold = get_onsets_by_cqt_rms(filename,onset_code)

if len(onset_frames_cqt) >= best_total:

best_total = len(onset_frames_cqt)

best_onset_frames_cqt = onset_frames_cqt

best_threshold = threshold

if len(onset_frames_cqt) < topN and onsets_frames != get_real_onsets_frames_rhythm(y, modify_by_energy=True, gap=0.12):

onset_frames_cqt,threshold = get_onsets_by_cqt_rms(filename,onset_code, 0.12)

if len(onset_frames_cqt) >= best_total:

best_total = len(onset_frames_cqt)

best_onset_frames_cqt = onset_frames_cqt

best_threshold = threshold

if len(onset_frames_cqt) < topN and onsets_frames != get_real_onsets_frames_rhythm(y, modify_by_energy=True, gap=0.09):

onset_frames_cqt,threshold = get_onsets_by_cqt_rms(filename,onset_code, 0.09)

if len(onset_frames_cqt) >= best_total:

best_total = len(onset_frames_cqt)

best_onset_frames_cqt = onset_frames_cqt

best_threshold = threshold

if len(best_onset_frames_cqt) <1:

best_onset_frames_cqt = onsets_frames

type_index = get_onsets_index_by_filename(filename)

total_frames_number = get_total_frames_number(filename)

# base_frames = onsets_base_frames_rhythm(type_index,total_frames_number)

base_frames = onsets_base_frames(codes[type_index], total_frames_number)

# 标准节拍个数

topN = len(base_frames)

best_onset_frames_cqt = []

best_total = 0

best_threshold = 0

onset_frames_cqt,threshold = get_onsets_by_cqt_rms(filename)

if len(onset_frames_cqt) >= best_total:

best_onset_frames_cqt = onset_frames_cqt

best_threshold = threshold

y, sr = librosa.load(filename)

CQT = librosa.amplitude_to_db(librosa.cqt(y, sr=16000), ref=np.max)

onset_frames_cqt, best_threshold = get_onsets_by_cqt_rms_optimised(filename)

#print("5. onset_frames_cqt,best_threshold is {},{}".format(onset_frames_cqt, best_threshold))

# if len(onset_frames_cqt)<topN:

onset_frames_cqt = get_miss_onsets_by_cqt(y, onset_frames_cqt)

#print("6. onset_frames_cqt,best_threshold is {},{}".format(onset_frames_cqt, best_threshold))

#onset_frames_cqt = check_onset_by_cqt_v2(y, onset_frames_cqt)

#print("7. onset_frames_cqt,best_threshold is {},{}".format(onset_frames_cqt, best_threshold))

onset_frames_cqt_time = librosa.frames_to_time(onset_frames_cqt, sr=sr)

#print("onset_frames_cqt_time is {}".format(onset_frames_cqt_time))

type_index = get_onsets_index_by_filename(filename)

total_frames_number = get_total_frames_number(filename)

best_y = []

# 标准节拍时间点

if len(onset_frames_cqt)> 0:

base_frames = onsets_base_frames(codes[type_index], total_frames_number - onset_frames_cqt[0])

base_frames = [x + (onset_frames_cqt[0] - base_frames[0]) for x in base_frames]

min_d, best_y, onsets_frames = get_dtw_min(onset_frames_cqt, base_frames, 65)

else:

base_frames = onsets_base_frames(codes[type_index], total_frames_number)

base_onsets = librosa.frames_to_time(base_frames, sr=sr)

# librosa.display.specshow(CQT)

plt.figure(figsize=(10, 6))

plt.subplot(4, 1, 1) # 要生成两行两列，这是第一个图plt.subplot('行','列','编号')

# plt.colorbar(format='%+2.0f dB')

# plt.title('Constant-Q power spectrogram (note)')

librosa.display.specshow(CQT, y_axis='cqt_note', x_axis='time')

plt.vlines(onset_frames_cqt_time, 0, sr, color='y', linestyle='solid')

#plt.vlines(base_onsets, 0, sr, color='r', linestyle='solid')

# print(plt.figure)

plt.subplot(4, 1, 2) # 要生成两行两列，这是第一个图plt.subplot('行','列','编号')

librosa.display.waveplot(y, sr=sr)

plt.vlines(onset_frames_cqt_time, -1 * np.max(y), np.max(y), color='y', linestyle='solid')

plt.subplot(4, 1, 3)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

max_rms = np.max(rms)

# rms = np.diff(rms)

times = librosa.frames_to_time(np.arange(len(rms)))

# rms_on_onset_frames_cqt = [rms[x] for x in onset_frames_cqt]

# min_rms_on_onset_frames_cqt = np.min(rms_on_onset_frames_cqt)

# rms = [1 if x >=min_rms_on_onset_frames_cqt else 0 for x in rms]

plt.plot(times, rms)

# plt.axhline(min_rms_on_onset_frames_cqt)

plt.axhline(max_rms * best_threshold)

# plt.vlines(onsets_frames_rms_best_time, 0,np.max(rms), color='y', linestyle='solid')

plt.vlines(onset_frames_cqt_time, 0, np.max(rms), color='y', linestyle='solid')

#plt.vlines(base_onsets, 0, np.max(rms), color='r', linestyle='solid')

plt.xlim(0, np.max(times))

plt.subplot(4, 1, 4)

plt.vlines(base_onsets, 0, np.max(rms), color='r', linestyle='dashed')

plt.xlim(0, np.max(times))

plt.axhline(max_rms * best_threshold)

onset_frames_cqt, best_y, best_threshold, _ = get_detail_cqt_rms(filename)

y, sr = librosa.load(filename)

loss_frames = []

for i in range(len(onset_frames_cqt)-1):

start = onset_frames_cqt[i]

end = onset_frames_cqt[i+1]

if end - start > 30:

start_end_time = librosa.frames_to_time([start,end], sr=sr)

#print("start_end_time is {}".format(start_end_time))

y1,sr1 = librosa.load(filename,offset=start_end_time[0],duration=start_end_time[1] - start_end_time[0])

# 根据rms阀值线找漏的

if len(onset_frames_cqt) > 0:

threshold = 0.6

tmp = get_missing_by_best_threshod(y1, [start,end], threshold)

if len(tmp)>=3:

for j in range(1,len(tmp)-1):

loss_frames.append(tmp[j])

#print("add is {}".format(tmp[1:-1]))

# rms = librosa.feature.rmse(y=y1)[0]

# rms_on_onset_frames_cqt = [rms[x] for x in [start,end]]

# min_rms_on_onset_frames_cqt = np.min(rms_on_onset_frames_cqt)

# rms = [1 if x >=min_rms_on_onset_frames_cqt else 0 for x in rms]

#

# loss = [i for i in range(len(rms)-6) if rms[i] == 0 and rms[i+1] == 1 and np.min(rms[i+1:i+6]) == 1 and i < end and i > start ]

# for x in loss:

# loss_frames.append(x)

if len(loss_frames)>0:

for x in loss_frames:

onset_frames_cqt.append(x)

onset_frames_cqt.sort()

CQT = librosa.amplitude_to_db(librosa.cqt(y, sr=16000), ref=np.max)

#onset_frames_cqt, best_threshold = get_onsets_by_cqt_rms_optimised(filename)

#print("5. onset_frames_cqt,best_threshold is {},{}".format(onset_frames_cqt, best_threshold))

# if len(onset_frames_cqt)<topN:

onset_frames_cqt = get_miss_onsets_by_cqt(y, onset_frames_cqt)

onset_frames_cqt = find_false_onsets_rms_secondary_optimised(y, onset_frames_cqt, 0.1, 0.1)

if onset_frames_cqt:

min_width = 5

# print("min_width is {}".format(min_width))

onset_frames_cqt = del_overcrowding(onset_frames_cqt, min_width)

#print("6. onset_frames_cqt,best_threshold is {},{}".format(onset_frames_cqt, best_threshold))

#onset_frames_cqt = check_onset_by_cqt_v2(y, onset_frames_cqt)

#print("7. onset_frames_cqt,best_threshold is {},{}".format(onset_frames_cqt, best_threshold))

onset_frames_cqt_time = librosa.frames_to_time(onset_frames_cqt, sr=sr)

type_index = get_onsets_index_by_filename(filename)

total_frames_number = get_total_frames_number(filename)

best_y = []

# 标准节拍时间点

if len(onset_frames_cqt)> 0:

base_frames = onsets_base_frames_for_note(filename)

base_frames = [x + onset_frames_cqt[0] - base_frames[0] for x in base_frames]

min_d, best_y, onsets_frames = get_dtw_min(onset_frames_cqt, base_frames, 65)

else:

base_frames = onsets_base_frames_for_note(filename)

base_onsets = librosa.frames_to_time(base_frames, sr=sr)

plt.close() # 关闭第一次的图片句柄

# librosa.display.specshow(CQT)

plt.figure(figsize=(10, 6))

plt.subplot(4, 1, 1) # 要生成两行两列，这是第一个图plt.subplot('行','列','编号')

# plt.colorbar(format='%+2.0f dB')

# plt.title('Constant-Q power spectrogram (note)')

# for x in onset_frames_cqt:

# sub_cqt = CQT.copy()[:,x]

# sub_cqt[0:20] = np.min(CQT)

# max_index = np.where(sub_cqt==np.max(sub_cqt))[0][0]

# print("max_index is {}".format(max_index))

# #plt.axhline(max_index,color="r")

# CQT[max_index,:] = np.min(CQT)

librosa.display.specshow(CQT, y_axis='cqt_note', x_axis='time')

plt.vlines(onset_frames_cqt_time, 0, sr, color='y', linestyle='solid')

#plt.vlines(base_onsets, 0, sr, color='r', linestyle='solid')

# print(plt.figure)

plt.subplot(4, 1, 2) # 要生成两行两列，这是第一个图plt.subplot('行','列','编号')

librosa.display.waveplot(y, sr=sr)

plt.vlines(onset_frames_cqt_time, -1 * np.max(y), np.max(y), color='y', linestyle='solid')

plt.subplot(4, 1, 3)

rms = librosa.feature.rmse(y=y)[0]

rms = [x / np.std(rms) for x in rms]

max_rms = np.max(rms)

# rms = np.diff(rms)

times = librosa.frames_to_time(np.arange(len(rms)))

rms_on_onset_frames_cqt = [rms[x] for x in onset_frames_cqt]

min_rms_on_onset_frames_cqt = np.min(rms_on_onset_frames_cqt)

rms = [1 if x >=min_rms_on_onset_frames_cqt else 0 for x in rms]

plt.plot(times, rms)

# plt.axhline(min_rms_on_onset_frames_cqt)

plt.axhline(max_rms * best_threshold)

# plt.vlines(onsets_frames_rms_best_time, 0,np.max(rms), color='y', linestyle='solid')

plt.vlines(onset_frames_cqt_time, 0, np.max(rms), color='y', linestyle='solid')

#plt.vlines(base_onsets, 0, np.max(rms), color='r', linestyle='solid')

plt.xlim(0, np.max(times))

plt.subplot(4, 1, 4)

plt.vlines(base_onsets, 0, np.max(rms), color='r', linestyle='dashed')

plt.xlim(0, np.max(times))

plt.axhline(max_rms * best_threshold)