1453 lines
60 KiB
C#
1453 lines
60 KiB
C#
using System;
|
||
using System.Collections.Concurrent;
|
||
using System.IO;
|
||
using System.Net.WebSockets;
|
||
using System.Security.Cryptography;
|
||
using System.Text;
|
||
using System.Threading;
|
||
using System.Threading.Tasks;
|
||
using Microsoft.Extensions.Logging;
|
||
using Newtonsoft.Json;
|
||
using ShengShengBuXi.Models;
|
||
using System.Collections.Generic;
|
||
using System.Linq;
|
||
using System.Net;
|
||
using System.Web;
|
||
|
||
namespace ShengShengBuXi.Services
|
||
{
|
||
/// <summary>
|
||
/// 语音识别会话
|
||
/// </summary>
|
||
public class SpeechRecognitionSession
|
||
{
|
||
/// <summary>
|
||
/// 会话ID
|
||
/// </summary>
|
||
public string SessionId { get; set; }
|
||
|
||
/// <summary>
|
||
/// 会话是否存在
|
||
/// </summary>
|
||
public bool Exists { get; set; }
|
||
|
||
/// <summary>
|
||
/// 会话是否活跃(WebSocket连接打开)
|
||
/// </summary>
|
||
public bool IsActive { get; set; }
|
||
|
||
/// <summary>
|
||
/// 是否已保存识别文本
|
||
/// </summary>
|
||
public bool HasSavedText { get; set; }
|
||
|
||
/// <summary>
|
||
/// 是否有待保存的文本
|
||
/// </summary>
|
||
public bool HasPendingText { get; set; }
|
||
|
||
/// <summary>
|
||
/// 待保存的文本内容
|
||
/// </summary>
|
||
public string PendingText { get; set; }
|
||
|
||
/// <summary>
|
||
/// 最终识别文本
|
||
/// </summary>
|
||
public string FinalText { get; set; }
|
||
|
||
/// <summary>
|
||
/// 录音文件路径
|
||
/// </summary>
|
||
public string RecordingPath { get; set; }
|
||
/// <summary>
|
||
/// 识别文字的文件的路径
|
||
/// </summary>
|
||
|
||
public string TextFilePath { get; set; }
|
||
|
||
/// <summary>
|
||
/// 是否已发送到大屏队列
|
||
/// </summary>
|
||
public bool IsSentToDisplay { get; set; }
|
||
|
||
/// <summary>
|
||
/// 采样率
|
||
/// </summary>
|
||
public int SampleRate { get; set; } = 16000;
|
||
|
||
/// <summary>
|
||
/// 语言代码
|
||
/// </summary>
|
||
public string Language { get; set; } = "zh-CN";
|
||
|
||
/// <summary>
|
||
/// 识别的文本内容列表
|
||
/// </summary>
|
||
public List<string> RecognizedTexts { get; set; } = new List<string>();
|
||
|
||
/// <summary>
|
||
/// 临时语音识别文本
|
||
/// </summary>
|
||
public string VoiceTextStr { get; set; }
|
||
|
||
/// <summary>
|
||
/// 会话创建时间
|
||
/// </summary>
|
||
public DateTime CreatedTime { get; set; } = DateTime.Now;
|
||
|
||
/// <summary>
|
||
/// 最后活动时间
|
||
/// </summary>
|
||
public DateTime LastActivityTime { get; set; } = DateTime.Now;
|
||
}
|
||
|
||
/// <summary>
|
||
/// 语音识别服务实现 - 腾讯云实时语音识别
|
||
/// </summary>
|
||
public class SpeechToTextService : ISpeechToTextService
|
||
{
|
||
private readonly ILogger<SpeechToTextService> _logger;
|
||
private readonly IConfigurationService _configService;
|
||
|
||
// 用于存储各个会话的WebSocket客户端
|
||
private static ConcurrentDictionary<string, ClientWebSocket> _webSockets = new ConcurrentDictionary<string, ClientWebSocket>();
|
||
|
||
// 用于存储各个会话的取消令牌
|
||
private readonly ConcurrentDictionary<string, CancellationTokenSource> _cancellationTokens = new ConcurrentDictionary<string, CancellationTokenSource>();
|
||
|
||
// 存储会话状态信息
|
||
private readonly ConcurrentDictionary<string, SpeechRecognitionSession> _sessions = new ConcurrentDictionary<string, SpeechRecognitionSession>();
|
||
|
||
private bool _isDisposed;
|
||
|
||
/// <summary>
|
||
/// 当有新的语音转文字结果时触发
|
||
/// </summary>
|
||
public event EventHandler<SpeechToTextResult> ResultReceived;
|
||
|
||
/// <summary>
|
||
/// 初始化语音识别服务
|
||
/// </summary>
|
||
/// <param name="logger">日志记录器</param>
|
||
/// <param name="configService">配置服务</param>
|
||
public SpeechToTextService(ILogger<SpeechToTextService> logger, IConfigurationService configService)
|
||
{
|
||
_logger = logger ?? throw new ArgumentNullException(nameof(logger));
|
||
_configService = configService ?? throw new ArgumentNullException(nameof(configService));
|
||
}
|
||
|
||
/// <summary>
|
||
/// 初始化语音识别服务
|
||
/// </summary>
|
||
/// <returns>初始化是否成功</returns>
|
||
public bool Initialize()
|
||
{
|
||
try
|
||
{
|
||
_logger.LogInformation("腾讯云语音识别服务初始化");
|
||
|
||
// 检查腾讯云配置是否有效
|
||
var config = _configService.CurrentConfig?.Network?.TencentCloudASR;
|
||
if (config == null || string.IsNullOrEmpty(config.AppId) ||
|
||
string.IsNullOrEmpty(config.SecretId) || string.IsNullOrEmpty(config.SecretKey))
|
||
{
|
||
_logger.LogWarning("腾讯云语音识别配置无效，请检查配置");
|
||
_logger.LogInformation($"腾讯云配置 - AppId: '{config?.AppId ?? "null"}', SecretId: '{(string.IsNullOrEmpty(config?.SecretId) ? "未设置" : "已设置")}', SecretKey: '{(string.IsNullOrEmpty(config?.SecretKey) ? "未设置" : "已设置")}'");
|
||
return false;
|
||
}
|
||
|
||
_logger.LogInformation($"腾讯云配置 - AppId: '{config.AppId}', SecretId: '{(string.IsNullOrEmpty(config.SecretId) ? "未设置" : "已设置")}', SecretKey: '{(string.IsNullOrEmpty(config.SecretKey) ? "未设置" : "已设置")}'");
|
||
|
||
return true;
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"语音识别服务初始化失败: {ex.Message}");
|
||
return false;
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 开始新的识别会话
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <param name="sampleRate">采样率</param>
|
||
/// <param name="language">语言代码</param>
|
||
/// <param name="token">取消令牌</param>
|
||
/// <returns>异步任务</returns>
|
||
public async Task StartSessionAsync(string sessionId, int sampleRate = 16000, string language = "zh-CN", CancellationToken token = default)
|
||
{
|
||
if (string.IsNullOrEmpty(sessionId))
|
||
{
|
||
throw new ArgumentNullException(nameof(sessionId));
|
||
}
|
||
|
||
// 结束可能存在的旧会话
|
||
await EndSessionAsync(sessionId, token);
|
||
|
||
try
|
||
{
|
||
// 检查配置是否有效
|
||
if (!Initialize())
|
||
{
|
||
_logger.LogError("腾讯云语音识别服务初始化失败，请检查配置");
|
||
throw new InvalidOperationException("语音识别服务未正确初始化，无法启动会话");
|
||
}
|
||
|
||
// 创建新会话
|
||
var session = new SpeechRecognitionSession
|
||
{
|
||
SessionId = sessionId,
|
||
SampleRate = sampleRate,
|
||
Language = language,
|
||
CreatedTime = DateTime.Now,
|
||
LastActivityTime = DateTime.Now,
|
||
IsActive = true,
|
||
Exists = true
|
||
};
|
||
|
||
_sessions[sessionId] = session;
|
||
|
||
// 创建取消令牌
|
||
var cts = new CancellationTokenSource();
|
||
_cancellationTokens[sessionId] = cts;
|
||
|
||
// 使用链接的取消令牌
|
||
var linkedCts = CancellationTokenSource.CreateLinkedTokenSource(cts.Token, token);
|
||
|
||
// 创建新的WebSocket客户端
|
||
var webSocket = new ClientWebSocket();
|
||
_webSockets[sessionId] = webSocket;
|
||
|
||
try
|
||
{
|
||
// 构建腾讯云实时语音识别WebSocket URL
|
||
string url = BuildTencentCloudASRUrl(sessionId);
|
||
|
||
// 连接到腾讯云WebSocket服务器
|
||
await webSocket.ConnectAsync(new Uri(url), linkedCts.Token);
|
||
_logger.LogInformation($"腾讯云WebSocket连接成功: {sessionId}");
|
||
|
||
_logger.LogInformation($"开始语音识别会话: {sessionId}, 采样率: {sampleRate}, 语言: {language}");
|
||
|
||
// 在后台接收识别结果
|
||
_ = ReceiveResultsAsync(sessionId, webSocket, linkedCts.Token);
|
||
}
|
||
catch (WebSocketException wsEx)
|
||
{
|
||
_logger.LogError($"WebSocket连接失败: {wsEx.Message}, WebSocketError: {wsEx.WebSocketErrorCode}");
|
||
throw;
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"连接腾讯云服务失败: {ex.Message}");
|
||
throw;
|
||
}
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"开始语音识别会话失败: {ex.Message}");
|
||
await EndSessionAsync(sessionId, token);
|
||
throw;
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 结束识别会话
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <param name="token">取消令牌</param>
|
||
/// <returns>异步任务</returns>
|
||
public async Task EndSessionAsync(string sessionId, CancellationToken token = default)
|
||
{
|
||
if (string.IsNullOrEmpty(sessionId))
|
||
{
|
||
throw new ArgumentNullException(nameof(sessionId));
|
||
}
|
||
|
||
_logger.LogInformation($"结束语音识别会话: {sessionId}");
|
||
|
||
// 获取会话
|
||
if (_sessions.TryGetValue(sessionId, out var session) && (session.RecognizedTexts.Count > 0 || session.VoiceTextStr?.Length > 0))
|
||
{
|
||
//
|
||
if (session.RecognizedTexts.Count == 0)
|
||
{
|
||
session.RecognizedTexts.Add(session.VoiceTextStr ?? "");
|
||
}
|
||
// 添加去重处理
|
||
var uniqueTexts = RemoveDuplicateTexts(session.RecognizedTexts);
|
||
string finalText = string.Join("\n", uniqueTexts);
|
||
_logger.LogInformation($"在结束会话前保存会话 {sessionId} 的所有已收集文本");
|
||
SaveRecognitionText(sessionId, finalText);
|
||
|
||
// 更新会话状态
|
||
session.IsActive = false;
|
||
session.LastActivityTime = DateTime.Now;
|
||
_sessions[sessionId] = session;
|
||
}
|
||
|
||
// 获取会话的WebSocket客户端
|
||
if (_webSockets.TryRemove(sessionId, out var webSocket))
|
||
{
|
||
try
|
||
{
|
||
// 发送结束消息
|
||
if (webSocket.State == WebSocketState.Open)
|
||
{
|
||
try
|
||
{
|
||
var endMessage = Encoding.UTF8.GetBytes("{\"type\": \"end\"}");
|
||
await webSocket.SendAsync(new ArraySegment<byte>(endMessage), WebSocketMessageType.Text, true, token);
|
||
|
||
// 等待片刻以接收最终结果
|
||
await Task.Delay(500, token);
|
||
}
|
||
catch (WebSocketException wsEx) when (wsEx.Message.Contains("remote party closed") ||
|
||
wsEx.Message.Contains("connection was aborted") ||
|
||
wsEx.Message.Contains("connection has been terminated"))
|
||
{
|
||
// 腾讯云服务器已关闭连接，无需发送结束消息
|
||
_logger.LogDebug($"发送结束消息时发现WebSocket已关闭: {sessionId}");
|
||
}
|
||
}
|
||
|
||
// 关闭WebSocket连接
|
||
if (webSocket.State == WebSocketState.Open)
|
||
{
|
||
try
|
||
{
|
||
await webSocket.CloseAsync(WebSocketCloseStatus.NormalClosure, "Session ended", token);
|
||
}
|
||
catch (WebSocketException wsEx) when (wsEx.Message.Contains("remote party closed") ||
|
||
wsEx.Message.Contains("connection was aborted") ||
|
||
wsEx.Message.Contains("connection has been terminated"))
|
||
{
|
||
// 腾讯云服务器已关闭连接，无需发送关闭帧
|
||
_logger.LogDebug($"关闭WebSocket时发现连接已关闭: {sessionId}");
|
||
}
|
||
}
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogWarning($"关闭WebSocket连接失败: {ex.Message}");
|
||
}
|
||
finally
|
||
{
|
||
// 无论成功与否，都释放资源
|
||
webSocket.Dispose();
|
||
}
|
||
}
|
||
|
||
// 取消后台任务
|
||
if (_cancellationTokens.TryRemove(sessionId, out var cts))
|
||
{
|
||
try
|
||
{
|
||
cts.Cancel();
|
||
cts.Dispose();
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogWarning($"取消任务失败: {ex.Message}");
|
||
}
|
||
}
|
||
|
||
// 注意：不要完全移除会话，因为可能还需要会话信息
|
||
if (_sessions.TryGetValue(sessionId, out session))
|
||
{
|
||
session.IsActive = false;
|
||
_sessions[sessionId] = session;
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 处理音频数据
|
||
/// </summary>
|
||
/// <param name="audioData">音频数据</param>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <param name="token">取消令牌</param>
|
||
/// <returns>异步任务</returns>
|
||
public async Task ProcessAudioAsync(byte[] audioData, string sessionId, CancellationToken token = default)
|
||
{
|
||
if (audioData == null || audioData.Length == 0)
|
||
{
|
||
_logger.LogWarning("音频数据为空");
|
||
return;
|
||
}
|
||
|
||
if (string.IsNullOrEmpty(sessionId))
|
||
{
|
||
_logger.LogWarning("会话Id为空");
|
||
throw new ArgumentNullException(nameof(sessionId));
|
||
}
|
||
|
||
// 检查会话是否存在
|
||
if (!_webSockets.TryGetValue(sessionId, out var webSocket))
|
||
{
|
||
_logger.LogWarning($"ProcessAudioAsync->会话不存在: {sessionId}");
|
||
return;
|
||
}
|
||
|
||
// 更新会话最后活动时间
|
||
if (_sessions.TryGetValue(sessionId, out var session))
|
||
{
|
||
session.LastActivityTime = DateTime.Now;
|
||
_sessions[sessionId] = session;
|
||
}
|
||
|
||
// 检查WebSocket状态
|
||
if (webSocket.State != WebSocketState.Open)
|
||
{
|
||
_logger.LogWarning($"WebSocket连接已关闭，尝试重新连接: {sessionId}");
|
||
|
||
// 获取会话信息
|
||
int sampleRate = 16000;
|
||
string language = "zh-CN";
|
||
|
||
if (_sessions.TryGetValue(sessionId, out session))
|
||
{
|
||
sampleRate = session.SampleRate;
|
||
language = session.Language;
|
||
}
|
||
|
||
// 重新开始会话
|
||
try
|
||
{
|
||
await StartSessionAsync(sessionId, sampleRate, language, token);
|
||
|
||
// 重新获取WebSocket
|
||
if (!_webSockets.TryGetValue(sessionId, out webSocket) || webSocket.State != WebSocketState.Open)
|
||
{
|
||
_logger.LogError($"重新连接失败，无法处理音频数据: {sessionId}");
|
||
return;
|
||
}
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"重新连接失败: {ex.Message}");
|
||
return;
|
||
}
|
||
}
|
||
|
||
try
|
||
{
|
||
// 将音频数据发送到腾讯云WebSocket服务器
|
||
await webSocket.SendAsync(new ArraySegment<byte>(audioData), WebSocketMessageType.Binary, true, token);
|
||
}
|
||
catch (WebSocketException wsEx) when (wsEx.Message.Contains("remote party closed") ||
|
||
wsEx.Message.Contains("connection was aborted") ||
|
||
wsEx.Message.Contains("connection has been terminated"))
|
||
{
|
||
_logger.LogWarning($"发送音频数据时WebSocket已关闭，尝试重新连接: {sessionId}");
|
||
|
||
// 获取会话信息
|
||
int sampleRate = 16000;
|
||
string language = "zh-CN";
|
||
|
||
if (_sessions.TryGetValue(sessionId, out session))
|
||
{
|
||
sampleRate = session.SampleRate;
|
||
language = session.Language;
|
||
}
|
||
|
||
// 重新开始会话
|
||
try
|
||
{
|
||
await StartSessionAsync(sessionId, sampleRate, language, token);
|
||
|
||
// 重试发送音频数据
|
||
if (_webSockets.TryGetValue(sessionId, out var newWebSocket) && newWebSocket.State == WebSocketState.Open)
|
||
{
|
||
await newWebSocket.SendAsync(new ArraySegment<byte>(audioData), WebSocketMessageType.Binary, true, token);
|
||
}
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"重新连接并发送数据失败: {ex.Message}");
|
||
}
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"发送音频数据失败: {ex.Message}");
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 处理音频文件
|
||
/// </summary>
|
||
/// <param name="audioFile">音频文件路径</param>
|
||
/// <param name="token">取消令牌</param>
|
||
/// <returns>识别结果</returns>
|
||
public async Task<SpeechToTextResult> ProcessAudioFileAsync(string audioFile, CancellationToken token = default)
|
||
{
|
||
if (string.IsNullOrEmpty(audioFile) || !File.Exists(audioFile))
|
||
{
|
||
throw new FileNotFoundException("音频文件不存在", audioFile);
|
||
}
|
||
|
||
// 创建一个临时会话ID
|
||
string sessionId = Guid.NewGuid().ToString();
|
||
|
||
try
|
||
{
|
||
// 开始会话
|
||
await StartSessionAsync(sessionId, 16000, "zh-CN", token);
|
||
|
||
// 读取音频文件
|
||
byte[] audioData = await File.ReadAllBytesAsync(audioFile, token);
|
||
|
||
// 这里可能需要根据文件格式进行转换，以匹配腾讯云API要求的格式
|
||
// 当前假设文件已经是正确的PCM格式
|
||
|
||
// 处理音频数据
|
||
await ProcessAudioAsync(audioData, sessionId, token);
|
||
|
||
// 结束会话
|
||
await EndSessionAsync(sessionId, token);
|
||
|
||
// 返回一个空的结果，因为实际结果会通过事件发送
|
||
return new SpeechToTextResult
|
||
{
|
||
Id = sessionId,
|
||
Text = "音频文件处理完成，结果将通过事件发送",
|
||
IsFinal = true,
|
||
RecordingFileName = Path.GetFileName(audioFile)
|
||
};
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"处理音频文件失败: {ex.Message}");
|
||
|
||
// 确保会话已结束
|
||
await EndSessionAsync(sessionId, token);
|
||
|
||
throw;
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 处理音频流
|
||
/// </summary>
|
||
/// <param name="audioStream">音频流</param>
|
||
/// <param name="token">取消令牌</param>
|
||
/// <returns>识别结果</returns>
|
||
public async Task<SpeechToTextResult> ProcessAudioStreamAsync(Stream audioStream, CancellationToken token = default)
|
||
{
|
||
if (audioStream == null)
|
||
{
|
||
throw new ArgumentNullException(nameof(audioStream));
|
||
}
|
||
|
||
// 创建一个临时会话ID
|
||
string sessionId = Guid.NewGuid().ToString();
|
||
|
||
try
|
||
{
|
||
// 开始会话
|
||
await StartSessionAsync(sessionId, 16000, "zh-CN", token);
|
||
|
||
// 读取音频流
|
||
byte[] buffer = new byte[16000]; // 大约1秒的音频数据（16000Hz，16位，单声道）
|
||
int bytesRead;
|
||
|
||
while ((bytesRead = await audioStream.ReadAsync(buffer, 0, buffer.Length, token)) > 0)
|
||
{
|
||
// 创建一个正确大小的数组
|
||
byte[] audioData = new byte[bytesRead];
|
||
Array.Copy(buffer, audioData, bytesRead);
|
||
|
||
// 处理音频数据
|
||
await ProcessAudioAsync(audioData, sessionId, token);
|
||
|
||
// 等待一段时间，模拟实时处理
|
||
await Task.Delay(100, token);
|
||
}
|
||
|
||
// 结束会话
|
||
await EndSessionAsync(sessionId, token);
|
||
|
||
// 返回一个空的结果，因为实际结果会通过事件发送
|
||
return new SpeechToTextResult
|
||
{
|
||
Id = sessionId,
|
||
Text = "音频流处理完成，结果将通过事件发送",
|
||
IsFinal = true
|
||
};
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"处理音频流失败: {ex.Message}");
|
||
|
||
// 确保会话已结束
|
||
await EndSessionAsync(sessionId, token);
|
||
|
||
throw;
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 设置会话的录音文件路径
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <param name="recordingPath">录音文件路径</param>
|
||
public void SetSessionRecordingPath(string sessionId, string recordingPath)
|
||
{
|
||
if (string.IsNullOrEmpty(sessionId) || string.IsNullOrEmpty(recordingPath))
|
||
return;
|
||
|
||
if (_sessions.TryGetValue(sessionId, out var session))
|
||
{
|
||
session.RecordingPath = recordingPath;
|
||
_sessions[sessionId] = session;
|
||
}
|
||
else
|
||
{
|
||
// 如果会话不存在，创建一个新的
|
||
session = new SpeechRecognitionSession
|
||
{
|
||
SessionId = sessionId,
|
||
RecordingPath = recordingPath,
|
||
Exists = true,
|
||
CreatedTime = DateTime.Now,
|
||
LastActivityTime = DateTime.Now
|
||
};
|
||
_sessions[sessionId] = session;
|
||
}
|
||
|
||
_logger.LogInformation($"设置会话 {sessionId} 的录音文件路径: {recordingPath}");
|
||
}
|
||
|
||
/// <summary>
|
||
/// 将识别文本保存到文件
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <param name="finalText">最终文本</param>
|
||
private void SaveRecognitionText(string sessionId, string finalText)
|
||
{
|
||
if (string.IsNullOrEmpty(sessionId) || string.IsNullOrEmpty(finalText))
|
||
return;
|
||
|
||
try
|
||
{
|
||
// 获取会话
|
||
if (!_sessions.TryGetValue(sessionId, out var session))
|
||
{
|
||
session = new SpeechRecognitionSession
|
||
{
|
||
SessionId = sessionId,
|
||
Exists = true,
|
||
CreatedTime = DateTime.Now,
|
||
LastActivityTime = DateTime.Now
|
||
};
|
||
_sessions[sessionId] = session;
|
||
}
|
||
|
||
// 检查录音文件路径
|
||
if (string.IsNullOrEmpty(session.RecordingPath))
|
||
{
|
||
_logger.LogWarning($"未找到会话 {sessionId} 的录音文件路径，无法保存文本");
|
||
|
||
// 将文本保存到会话中
|
||
session.PendingText = finalText;
|
||
session.HasPendingText = true;
|
||
_sessions[sessionId] = session;
|
||
|
||
// 启动后台任务，定期检查是否可以保存
|
||
StartPendingTextSaver(sessionId);
|
||
|
||
return;
|
||
}
|
||
|
||
// 生成文本文件路径 (与录音文件同名，但扩展名为.txt)
|
||
string textFilePath = Path.ChangeExtension(session.RecordingPath, ".txt");
|
||
session.TextFilePath = textFilePath;
|
||
// 确保目录存在
|
||
string directory = Path.GetDirectoryName(textFilePath);
|
||
if (!string.IsNullOrEmpty(directory) && !Directory.Exists(directory))
|
||
{
|
||
Directory.CreateDirectory(directory);
|
||
}
|
||
|
||
// 保存文本到文件
|
||
int retryCount = 0;
|
||
const int maxRetries = 3;
|
||
bool saved = false;
|
||
|
||
while (!saved && retryCount < maxRetries)
|
||
{
|
||
try
|
||
{
|
||
File.WriteAllText(textFilePath, finalText);
|
||
saved = true;
|
||
_logger.LogInformation($"保存识别文本到文件 {textFilePath}");
|
||
|
||
// 更新会话状态
|
||
session.HasSavedText = true;
|
||
session.HasPendingText = false;
|
||
session.PendingText = null;
|
||
session.FinalText = finalText;
|
||
_sessions[sessionId] = session;
|
||
}
|
||
catch (IOException ioEx)
|
||
{
|
||
// 文件可能被占用，等待一段时间后重试
|
||
retryCount++;
|
||
_logger.LogWarning($"保存文本失败 (尝试 {retryCount}/{maxRetries}): {ioEx.Message}");
|
||
|
||
if (retryCount < maxRetries)
|
||
{
|
||
Thread.Sleep(500 * retryCount); // 逐渐增加等待时间
|
||
}
|
||
}
|
||
}
|
||
|
||
if (!saved)
|
||
{
|
||
// 如果重试后仍然失败，将文本保存到备用文件
|
||
string backupPath = Path.Combine(
|
||
Path.GetDirectoryName(textFilePath) ?? string.Empty,
|
||
$"backup_{Path.GetFileNameWithoutExtension(textFilePath)}_{DateTime.Now:yyyyMMdd_HHmmss}.txt"
|
||
);
|
||
|
||
try
|
||
{
|
||
File.WriteAllText(backupPath, finalText);
|
||
_logger.LogInformation($"已将识别文本保存到备用文件: {backupPath}");
|
||
}
|
||
catch (Exception backupEx)
|
||
{
|
||
_logger.LogError($"保存识别文本到备用文件失败: {backupEx.Message}");
|
||
}
|
||
}
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"保存识别文本失败: {ex.Message}");
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 启动后台任务检查和保存等待的文本
|
||
/// </summary>
|
||
private void StartPendingTextSaver(string sessionId)
|
||
{
|
||
Task.Run(async () =>
|
||
{
|
||
// 最多等待30秒
|
||
for (int i = 0; i < 15; i++)
|
||
{
|
||
// 检查会话是否存在，是否有待保存文本
|
||
if (!_sessions.TryGetValue(sessionId, out var session) || !session.HasPendingText)
|
||
{
|
||
return;
|
||
}
|
||
|
||
// 检查是否已设置录音路径
|
||
if (!string.IsNullOrEmpty(session.RecordingPath))
|
||
{
|
||
_logger.LogInformation($"发现会话 {sessionId} 的录音文件路径，准备保存之前缓存的文本");
|
||
|
||
// 生成文本文件路径
|
||
string textFilePath = Path.ChangeExtension(session.RecordingPath, ".txt");
|
||
try
|
||
{
|
||
// 确保目录存在
|
||
string directory = Path.GetDirectoryName(textFilePath);
|
||
if (!string.IsNullOrEmpty(directory) && !Directory.Exists(directory))
|
||
{
|
||
Directory.CreateDirectory(directory);
|
||
}
|
||
session.TextFilePath = textFilePath;
|
||
// 保存文本
|
||
File.WriteAllText(textFilePath, session.PendingText);
|
||
_logger.LogInformation($"成功保存延迟的识别文本到文件 {textFilePath}");
|
||
|
||
// 更新会话状态
|
||
session.HasSavedText = true;
|
||
session.HasPendingText = false;
|
||
session.PendingText = null;
|
||
_sessions[sessionId] = session;
|
||
|
||
return;
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"保存延迟识别文本失败: {ex.Message}");
|
||
}
|
||
}
|
||
|
||
// 等待2秒后再次检查
|
||
await Task.Delay(2000);
|
||
}
|
||
|
||
// 如果超时后仍有未保存的文本，尝试保存到备用位置
|
||
if (_sessions.TryGetValue(sessionId, out var timeoutSession) && timeoutSession.HasPendingText)
|
||
{
|
||
string backupDir = Path.Combine(AppDomain.CurrentDomain.BaseDirectory, "backup_texts");
|
||
if (!Directory.Exists(backupDir))
|
||
{
|
||
Directory.CreateDirectory(backupDir);
|
||
}
|
||
|
||
string backupPath = Path.Combine(backupDir, $"session_{sessionId}_{DateTime.Now:yyyyMMdd_HHmmss}.txt");
|
||
timeoutSession.TextFilePath = backupPath;
|
||
try
|
||
{
|
||
File.WriteAllText(backupPath, timeoutSession.PendingText);
|
||
_logger.LogInformation($"已将超时未保存的识别文本保存到备用位置: {backupPath}");
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"保存超时识别文本到备用位置失败: {ex.Message}");
|
||
}
|
||
}
|
||
});
|
||
}
|
||
|
||
/// <summary>
|
||
/// 接收识别结果
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <param name="webSocket">WebSocket客户端</param>
|
||
/// <param name="token">取消令牌</param>
|
||
/// <returns>异步任务</returns>
|
||
private async Task ReceiveResultsAsync(string sessionId, ClientWebSocket webSocket, CancellationToken token)
|
||
{
|
||
byte[] buffer = new byte[8192];
|
||
StringBuilder messageBuilder = new StringBuilder();
|
||
|
||
try
|
||
{
|
||
// 确保该会话存在
|
||
if (!_sessions.TryGetValue(sessionId, out var session))
|
||
{
|
||
session = new SpeechRecognitionSession
|
||
{
|
||
SessionId = sessionId,
|
||
Exists = true,
|
||
IsActive = true,
|
||
CreatedTime = DateTime.Now,
|
||
LastActivityTime = DateTime.Now
|
||
};
|
||
_sessions[sessionId] = session;
|
||
}
|
||
|
||
while (webSocket.State == WebSocketState.Open && !token.IsCancellationRequested)
|
||
{
|
||
WebSocketReceiveResult result = await webSocket.ReceiveAsync(new ArraySegment<byte>(buffer), token);
|
||
|
||
if (result.MessageType == WebSocketMessageType.Close)
|
||
{
|
||
await webSocket.CloseAsync(WebSocketCloseStatus.NormalClosure, "Response complete", token);
|
||
break;
|
||
}
|
||
|
||
if (result.MessageType == WebSocketMessageType.Text)
|
||
{
|
||
string message = Encoding.UTF8.GetString(buffer, 0, result.Count);
|
||
messageBuilder.Append(message);
|
||
|
||
if (result.EndOfMessage)
|
||
{
|
||
string fullMessage = messageBuilder.ToString();
|
||
messageBuilder.Clear();
|
||
|
||
_logger.LogDebug($"收到识别结果: {fullMessage}");
|
||
|
||
try
|
||
{
|
||
// 解析腾讯云返回的结果
|
||
var tencentResult = JsonConvert.DeserializeObject<TencentCloudASRResult>(fullMessage);
|
||
|
||
// 检查结果是否有效
|
||
if (tencentResult != null && tencentResult.Code == 0)
|
||
{
|
||
// 如果有结果数据
|
||
if (tencentResult.Result != null && !string.IsNullOrEmpty(tencentResult.Result.VoiceTextStr))
|
||
{
|
||
// 获取文本
|
||
string recognizedText = tencentResult.Result.VoiceTextStr;
|
||
|
||
// 重新获取最新的会话状态
|
||
if (_sessions.TryGetValue(sessionId, out session))
|
||
{
|
||
session.LastActivityTime = DateTime.Now;
|
||
|
||
if (tencentResult.Result.SliceType == 0 || tencentResult.Result.SliceType == 1)
|
||
{
|
||
// 中间结果，只用于显示，不加入最终保存列表
|
||
// 我们仍然触发事件通知UI更新，但标记为非最终结果
|
||
var tempResult = new SpeechToTextResult
|
||
{
|
||
Id = tencentResult.MessageId,
|
||
Text = recognizedText,
|
||
IsFinal = false,
|
||
Confidence = 1.0f,
|
||
Language = session.Language,
|
||
StartTimeMs = tencentResult.Result.StartTime,
|
||
EndTimeMs = tencentResult.Result.EndTime,
|
||
CreatedAt = DateTime.Now,
|
||
IsForDisplayOnly = true // 标记为仅用于显示的中间结果
|
||
};
|
||
session.VoiceTextStr = recognizedText;
|
||
// 触发临时结果事件
|
||
OnResultReceived(tempResult);
|
||
}
|
||
else if (tencentResult.Result.SliceType == 2)
|
||
{
|
||
// 一句话结束，这时才添加到最终保存列表
|
||
bool shouldAdd = true;
|
||
|
||
// 检查是否与已有文本重复
|
||
if (session.RecognizedTexts.Count > 0)
|
||
{
|
||
// 与最后一条比较相似度
|
||
string lastText = session.RecognizedTexts.Last();
|
||
double similarity = CalculateTextSimilarity(lastText, recognizedText);
|
||
|
||
// 如果相似度超过70%，则认为重复，只更新不添加
|
||
if (similarity > 0.7)
|
||
{
|
||
_logger.LogInformation($"检测到相似文本(相似度:{similarity:P2})，更新而不是添加");
|
||
session.RecognizedTexts[session.RecognizedTexts.Count - 1] = recognizedText; // 使用新的替换旧的
|
||
shouldAdd = false;
|
||
}
|
||
}
|
||
|
||
if (shouldAdd)
|
||
{
|
||
_logger.LogInformation($"句子完成，添加到保存列表: '{recognizedText}'");
|
||
session.RecognizedTexts.Add(recognizedText);
|
||
}
|
||
|
||
// 更新会话
|
||
_sessions[sessionId] = session;
|
||
|
||
// 每当一句话结束，立即保存当前累积的所有文本
|
||
// 这样可以确保每句话识别完毕后就保存一次
|
||
SaveLatestRecognitionText(sessionId, session.RecognizedTexts);
|
||
}
|
||
}
|
||
|
||
// 只对最终结果或句子结束触发完整事件
|
||
if (tencentResult.Final == 1 || tencentResult.Result.SliceType == 2)
|
||
{
|
||
var speechResult = new SpeechToTextResult
|
||
{
|
||
Id = tencentResult.MessageId,
|
||
Text = recognizedText,
|
||
IsFinal = true,
|
||
Confidence = 1.0f, // 腾讯云不返回置信度，默认为1.0
|
||
Language = session.Language,
|
||
StartTimeMs = tencentResult.Result.StartTime,
|
||
EndTimeMs = tencentResult.Result.EndTime,
|
||
CreatedAt = DateTime.Now,
|
||
RecordingFilePath = session.RecordingPath,
|
||
RecordingFileName = !string.IsNullOrEmpty(session.RecordingPath) ? Path.GetFileName(session.RecordingPath) : null
|
||
};
|
||
|
||
// 触发结果事件
|
||
OnResultReceived(speechResult);
|
||
}
|
||
}
|
||
|
||
// 如果是最终结果，结束会话并保存文本
|
||
if (tencentResult.Final == 1)
|
||
{
|
||
_logger.LogInformation($"识别完成，结束会话: {sessionId}");
|
||
|
||
// 合并所有文本并保存
|
||
if (_sessions.TryGetValue(sessionId, out session) && session.RecognizedTexts.Count > 0)
|
||
{
|
||
// 先进行文本去重处理
|
||
var uniqueTexts = RemoveDuplicateTexts(session.RecognizedTexts);
|
||
string finalText = string.Join("\n", uniqueTexts);
|
||
SaveRecognitionText(sessionId, finalText);
|
||
|
||
// 更新会话状态
|
||
session.IsActive = false;
|
||
_sessions[sessionId] = session;
|
||
}
|
||
}
|
||
}
|
||
else if (tencentResult != null && tencentResult.Code != 0)
|
||
{
|
||
_logger.LogWarning($"识别失败: {tencentResult.Code}, {tencentResult.Message}");
|
||
}
|
||
}
|
||
catch (JsonException ex)
|
||
{
|
||
_logger.LogError($"解析识别结果失败: {ex.Message}, 原始消息: {fullMessage}");
|
||
}
|
||
}
|
||
}
|
||
}
|
||
}
|
||
catch (WebSocketException wsEx) when (wsEx.Message.Contains("remote party closed") ||
|
||
wsEx.Message.Contains("connection was aborted") ||
|
||
wsEx.Message.Contains("connection has been terminated") ||
|
||
wsEx.Message.Contains("without completing the close handshake"))
|
||
{
|
||
// 腾讯云服务器直接关闭连接，这是正常现象
|
||
_logger.LogInformation($"腾讯云服务器已关闭连接，识别会话{sessionId}结束");
|
||
|
||
// 保存已收集的文本
|
||
if (_sessions.TryGetValue(sessionId, out var session) && session.RecognizedTexts.Count > 0)
|
||
{
|
||
// 先进行文本去重处理
|
||
var uniqueTexts = RemoveDuplicateTexts(session.RecognizedTexts);
|
||
string finalText = string.Join("\n", uniqueTexts);
|
||
SaveRecognitionText(sessionId, finalText);
|
||
|
||
// 更新会话状态
|
||
session.IsActive = false;
|
||
_sessions[sessionId] = session;
|
||
}
|
||
}
|
||
catch (OperationCanceledException)
|
||
{
|
||
_logger.LogInformation($"接收任务已取消: {sessionId}");
|
||
|
||
// 保存已收集的文本
|
||
if (_sessions.TryGetValue(sessionId, out var session) && session.RecognizedTexts.Count > 0)
|
||
{
|
||
// 先进行文本去重处理
|
||
var uniqueTexts = RemoveDuplicateTexts(session.RecognizedTexts);
|
||
string finalText = string.Join("\n", uniqueTexts);
|
||
SaveRecognitionText(sessionId, finalText);
|
||
|
||
// 更新会话状态
|
||
session.IsActive = false;
|
||
_sessions[sessionId] = session;
|
||
}
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"接收识别结果异常: {ex.Message}");
|
||
|
||
// 尝试保存已收集的文本
|
||
if (_sessions.TryGetValue(sessionId, out var session) && session.RecognizedTexts.Count > 0)
|
||
{
|
||
// 先进行文本去重处理
|
||
var uniqueTexts = RemoveDuplicateTexts(session.RecognizedTexts);
|
||
string finalText = string.Join("\n", uniqueTexts);
|
||
SaveRecognitionText(sessionId, finalText);
|
||
|
||
// 更新会话状态
|
||
session.IsActive = false;
|
||
_sessions[sessionId] = session;
|
||
}
|
||
}
|
||
finally
|
||
{
|
||
// 确保正常关闭WebSocket
|
||
if (webSocket.State == WebSocketState.Open)
|
||
{
|
||
try
|
||
{
|
||
// 使用一个新的取消令牌，避免使用已取消的令牌
|
||
await webSocket.CloseAsync(WebSocketCloseStatus.NormalClosure, "Receiving complete", CancellationToken.None);
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogDebug($"关闭WebSocket连接时发生异常: {ex.Message}");
|
||
}
|
||
}
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 计算两段文本的相似度
|
||
/// </summary>
|
||
/// <param name="text1">文本1</param>
|
||
/// <param name="text2">文本2</param>
|
||
/// <returns>相似度(0-1)</returns>
|
||
private double CalculateTextSimilarity(string text1, string text2)
|
||
{
|
||
if (string.IsNullOrEmpty(text1) || string.IsNullOrEmpty(text2))
|
||
return 0;
|
||
|
||
// 如果两段文本完全一样
|
||
if (text1 == text2)
|
||
return 1.0;
|
||
|
||
// 计算最长公共子串
|
||
int[,] dp = new int[text1.Length + 1, text2.Length + 1];
|
||
int maxLength = 0;
|
||
|
||
for (int i = 1; i <= text1.Length; i++)
|
||
{
|
||
for (int j = 1; j <= text2.Length; j++)
|
||
{
|
||
if (text1[i - 1] == text2[j - 1])
|
||
{
|
||
dp[i, j] = dp[i - 1, j - 1] + 1;
|
||
maxLength = Math.Max(maxLength, dp[i, j]);
|
||
}
|
||
}
|
||
}
|
||
|
||
// 计算相似度 (最长公共子串长度 / 较长文本的长度)
|
||
return (double)maxLength / Math.Max(text1.Length, text2.Length);
|
||
}
|
||
|
||
/// <summary>
|
||
/// 移除重复的文本
|
||
/// </summary>
|
||
/// <param name="texts">文本列表</param>
|
||
/// <returns>去重后的文本列表</returns>
|
||
private List<string> RemoveDuplicateTexts(List<string> texts)
|
||
{
|
||
if (texts == null || texts.Count <= 1)
|
||
return texts?.ToList() ?? new List<string>();
|
||
|
||
var result = new List<string>();
|
||
|
||
// 第一条总是添加
|
||
result.Add(texts[0]);
|
||
|
||
// 依次检查后面的文本
|
||
for (int i = 1; i < texts.Count; i++)
|
||
{
|
||
bool isDuplicate = false;
|
||
|
||
// 与已添加的所有文本比较
|
||
foreach (var existingText in result)
|
||
{
|
||
double similarity = CalculateTextSimilarity(existingText, texts[i]);
|
||
|
||
// 如果相似度超过70%，视为重复
|
||
if (similarity > 0.7)
|
||
{
|
||
isDuplicate = true;
|
||
_logger.LogInformation($"移除重复文本，相似度: {similarity:P2}");
|
||
break;
|
||
}
|
||
}
|
||
|
||
// 如果不是重复的，添加到结果中
|
||
if (!isDuplicate)
|
||
{
|
||
result.Add(texts[i]);
|
||
}
|
||
}
|
||
|
||
return result;
|
||
}
|
||
|
||
/// <summary>
|
||
/// 构建腾讯云实时语音识别WebSocket URL
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <returns>WebSocket URL</returns>
|
||
private string BuildTencentCloudASRUrl(string sessionId)
|
||
{
|
||
try
|
||
{
|
||
var config = _configService.CurrentConfig.Network.TencentCloudASR;
|
||
int sampleRate = 16000;
|
||
|
||
// 获取会话采样率
|
||
if (_sessions.TryGetValue(sessionId, out var session))
|
||
{
|
||
sampleRate = session.SampleRate;
|
||
}
|
||
|
||
// 检查关键参数是否为空
|
||
if (string.IsNullOrEmpty(config.AppId) ||
|
||
string.IsNullOrEmpty(config.SecretId) ||
|
||
string.IsNullOrEmpty(config.SecretKey))
|
||
{
|
||
_logger.LogError("腾讯云API参数缺失：AppId, SecretId 或 SecretKey 为空");
|
||
_logger.LogError($"当前配置 - AppId: '{config.AppId}', SecretId: '{(string.IsNullOrEmpty(config.SecretId) ? "未设置" : "已设置")}', SecretKey: '{(string.IsNullOrEmpty(config.SecretKey) ? "未设置" : "已设置")}'");
|
||
throw new InvalidOperationException("腾讯云API参数不能为空");
|
||
}
|
||
|
||
// 生成随机数（用作nonce）
|
||
var nonce = DateTimeOffset.UtcNow.ToUnixTimeSeconds().ToString();
|
||
|
||
// 当前时间戳
|
||
var timestamp = DateTimeOffset.UtcNow.ToUnixTimeSeconds().ToString();
|
||
|
||
// 过期时间，默认24小时
|
||
var expired = (DateTimeOffset.UtcNow.ToUnixTimeSeconds() + 86400).ToString();
|
||
|
||
// 语音ID，使用会话ID
|
||
var voiceId = sessionId;
|
||
|
||
// 根据采样率选择引擎模型
|
||
var engineModelType = sampleRate == 8000 ? "8k_zh" : "16k_zh";
|
||
if (!string.IsNullOrEmpty(config.EngineModelType))
|
||
{
|
||
engineModelType = config.EngineModelType;
|
||
}
|
||
|
||
// 构建参数字典
|
||
var parameters = new Dictionary<string, string>
|
||
{
|
||
{ "secretid", config.SecretId },
|
||
{ "timestamp", timestamp },
|
||
{ "expired", expired },
|
||
{ "nonce", nonce },
|
||
{ "voice_id", voiceId },
|
||
{ "voice_format", config.VoiceFormat }, // 1: pcm格式
|
||
{ "engine_model_type", engineModelType },
|
||
{ "needvad", config.NeedVad ? "1" : "0" }
|
||
};
|
||
|
||
// 可选参数
|
||
if (config.FilterDirty)
|
||
parameters.Add("filter_dirty", "1");
|
||
|
||
if (config.FilterModal)
|
||
parameters.Add("filter_modal", "1");
|
||
|
||
if (config.FilterPunc)
|
||
parameters.Add("filter_punc", "1");
|
||
|
||
// 构建签名原始字符串
|
||
var signatureOrigin = "asr.cloud.tencent.com/asr/v2/" + config.AppId + "?";
|
||
var paramList = new List<string>();
|
||
|
||
foreach (var param in parameters.OrderBy(p => p.Key))
|
||
{
|
||
paramList.Add($"{param.Key}={param.Value}");
|
||
}
|
||
|
||
signatureOrigin += string.Join("&", paramList);
|
||
|
||
// 使用HMAC-SHA1计算签名
|
||
using var hmac = new HMACSHA1(Encoding.UTF8.GetBytes(config.SecretKey));
|
||
var signatureBytes = hmac.ComputeHash(Encoding.UTF8.GetBytes(signatureOrigin));
|
||
var signature = Convert.ToBase64String(signatureBytes);
|
||
|
||
// URL编码签名
|
||
var encodedSignature = HttpUtility.UrlEncode(signature);
|
||
|
||
// 构建最终URL - 需要确保AppID包含在URL路径中
|
||
var url = $"wss://asr.cloud.tencent.com/asr/v2/{config.AppId}?";
|
||
url += string.Join("&", parameters.OrderBy(p => p.Key).Select(p => $"{p.Key}={HttpUtility.UrlEncode(p.Value)}")) + $"&signature={encodedSignature}";
|
||
|
||
_logger.LogInformation($"生成的签名原始字符串: {signatureOrigin}");
|
||
_logger.LogInformation($"生成的签名: {signature}");
|
||
_logger.LogInformation($"正在连接腾讯云WebSocket，URL: {url}");
|
||
|
||
return url;
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"构建腾讯云URL失败: {ex.Message}");
|
||
throw;
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 触发结果事件
|
||
/// </summary>
|
||
/// <param name="result">语音识别结果</param>
|
||
protected virtual void OnResultReceived(SpeechToTextResult result)
|
||
{
|
||
ResultReceived?.Invoke(this, result);
|
||
}
|
||
|
||
/// <summary>
|
||
/// 释放资源
|
||
/// </summary>
|
||
public void Dispose()
|
||
{
|
||
if (_isDisposed)
|
||
return;
|
||
|
||
_isDisposed = true;
|
||
|
||
// 结束所有会话并保存文本
|
||
foreach (var sessionId in _webSockets.Keys.ToList())
|
||
{
|
||
try
|
||
{
|
||
// 保存会话文本
|
||
if (_sessions.TryGetValue(sessionId, out var session) && session.RecognizedTexts.Count > 0)
|
||
{
|
||
string finalText = string.Join("\n", session.RecognizedTexts);
|
||
SaveRecognitionText(sessionId, finalText);
|
||
}
|
||
|
||
// 结束会话
|
||
EndSessionAsync(sessionId).Wait();
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogWarning($"释放会话资源失败: {ex.Message}");
|
||
}
|
||
}
|
||
|
||
// 清空所有集合
|
||
_webSockets.Clear();
|
||
_cancellationTokens.Clear();
|
||
_sessions.Clear();
|
||
}
|
||
|
||
/// <summary>
|
||
/// 保存最新的识别文本
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <param name="textList">文本列表</param>
|
||
private void SaveLatestRecognitionText(string sessionId, List<string> textList)
|
||
{
|
||
try
|
||
{
|
||
if (textList == null || textList.Count == 0)
|
||
return;
|
||
|
||
// 去重处理
|
||
var uniqueTexts = RemoveDuplicateTexts(textList);
|
||
string finalText = string.Join("\n", uniqueTexts);
|
||
|
||
// 保存文本
|
||
SaveRecognitionText(sessionId, finalText);
|
||
|
||
_logger.LogInformation($"已保存会话 {sessionId} 的最新识别文本，共 {uniqueTexts.Count} 句");
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"保存最新识别文本失败: {ex.Message}");
|
||
}
|
||
}
|
||
|
||
/// <summary>
|
||
/// 根据会话ID获取识别文字
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <returns>识别的文字，如果会话不存在则返回空字符串</returns>
|
||
public string GetRecognizedText(string sessionId)
|
||
{
|
||
if (string.IsNullOrEmpty(sessionId))
|
||
{
|
||
_logger.LogWarning("尝试获取识别文字时会话ID为空");
|
||
return string.Empty;
|
||
}
|
||
|
||
// 检查会话是否存在
|
||
if (_sessions.TryGetValue(sessionId, out var session))
|
||
{
|
||
// 检查会话中的识别文本
|
||
if (session.RecognizedTexts.Count > 0)
|
||
{
|
||
// 返回去重后的文本
|
||
var uniqueTexts = RemoveDuplicateTexts(session.RecognizedTexts);
|
||
return string.Join("\n", uniqueTexts);
|
||
}
|
||
|
||
// 检查是否有待保存的文本
|
||
if (session.HasPendingText && !string.IsNullOrEmpty(session.PendingText))
|
||
{
|
||
return session.PendingText;
|
||
}
|
||
|
||
// 如果会话已结束，尝试从文件中读取
|
||
if (!string.IsNullOrEmpty(session.RecordingPath))
|
||
{
|
||
string textFilePath = Path.ChangeExtension(session.RecordingPath, ".txt");
|
||
if (File.Exists(textFilePath))
|
||
{
|
||
try
|
||
{
|
||
return File.ReadAllText(textFilePath);
|
||
}
|
||
catch (Exception ex)
|
||
{
|
||
_logger.LogError($"从文件读取识别文本失败: {ex.Message}");
|
||
}
|
||
}
|
||
}
|
||
}
|
||
|
||
_logger.LogWarning($"未找到会话 {sessionId} 的识别文本");
|
||
return string.Empty;
|
||
}
|
||
|
||
/// <summary>
|
||
/// 获取会话状态信息
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <returns>会话状态信息</returns>
|
||
public SpeechRecognitionSession GetSessionStatus(string sessionId)
|
||
{
|
||
if (string.IsNullOrEmpty(sessionId))
|
||
{
|
||
_logger.LogWarning("尝试获取会话状态时会话ID为空");
|
||
return new SpeechRecognitionSession
|
||
{
|
||
SessionId = string.Empty,
|
||
Exists = false,
|
||
IsActive = false,
|
||
HasSavedText = false,
|
||
HasPendingText = false,
|
||
RecordingPath = string.Empty,
|
||
IsSentToDisplay = false
|
||
};
|
||
}
|
||
|
||
// 如果已存在状态信息，直接返回
|
||
if (_sessions.TryGetValue(sessionId, out var session))
|
||
{
|
||
// 确保状态信息是最新的
|
||
session.IsActive = _webSockets.ContainsKey(sessionId) &&
|
||
_webSockets[sessionId]?.State == WebSocketState.Open;
|
||
|
||
// 如果有录音文件路径，检查文本文件是否存在
|
||
if (!string.IsNullOrEmpty(session.RecordingPath))
|
||
{
|
||
string textFilePath = Path.ChangeExtension(session.RecordingPath, ".txt");
|
||
session.HasSavedText = File.Exists(textFilePath);
|
||
}
|
||
|
||
// 更新会话
|
||
_sessions[sessionId] = session;
|
||
|
||
return session;
|
||
}
|
||
|
||
// 创建新的状态信息
|
||
bool isActive = _webSockets.ContainsKey(sessionId) &&
|
||
_webSockets[sessionId]?.State == WebSocketState.Open;
|
||
|
||
session = new SpeechRecognitionSession
|
||
{
|
||
SessionId = sessionId,
|
||
Exists = isActive,
|
||
IsActive = isActive,
|
||
HasSavedText = false,
|
||
HasPendingText = false,
|
||
RecordingPath = string.Empty,
|
||
IsSentToDisplay = false,
|
||
CreatedTime = DateTime.Now,
|
||
LastActivityTime = DateTime.Now
|
||
};
|
||
|
||
// 保存状态信息
|
||
_sessions[sessionId] = session;
|
||
return session;
|
||
}
|
||
|
||
/// <summary>
|
||
/// 设置会话的大屏发送状态
|
||
/// </summary>
|
||
/// <param name="sessionId">会话ID</param>
|
||
/// <param name="isSent">是否已发送到大屏</param>
|
||
public void SetSessionDisplayStatus(string sessionId, bool isSent)
|
||
{
|
||
if (string.IsNullOrEmpty(sessionId))
|
||
return;
|
||
|
||
// 获取或创建会话状态
|
||
var session = GetSessionStatus(sessionId);
|
||
session.IsSentToDisplay = isSent;
|
||
_sessions[sessionId] = session;
|
||
|
||
_logger.LogInformation($"设置会话 {sessionId} 的大屏发送状态: {(isSent ? "已发送" : "未发送")}");
|
||
}
|
||
}
|
||
} |